www.sciencemag.org/cgi/content/full/science.aao1807/DC1

Supplementary Materials for

Ancient genomes show social and reproductive behavior of early Upper

Paleolithic foragers

Martin Sikora, Andaine Seguin-Orlando, Vitor C. Sousa, Anders Albrechtsen, Thorfinn Korneliussen, Amy Ko, Simon Rasmussen, Isabelle Dupanloup, Philip R. Nigst,

Marjolein D. Bosch, Gabriel Renaud, Morten E. Allentoft, Ashot Margaryan, Sergey V. Vasilyev, Elizaveta V. Veselovskaya, Svetlana B. Borutskaya, Thibaut Deviese,

Dan Comeskey, Tom Higham, Andrea Manica, Robert Foley, David J. Meltzer, Rasmus Nielsen, Laurent Excoffier, Marta Mirazon Lahr, Ludovic Orlando,

Eske Willerslev* *Corresponding author. Email: ewillerslev@snm.ku.dk

Published 5 October 2017 on Science First Release DOI: 10.1126/science.aao1807

This PDF file includes: Supplementary Text

Figs. S1 to S37

Tables S1 to S27

Captions for tables S28 and S29

References

Other supplementary material for this manuscript includes the following:

Tables S28 and S29 (Excel format)

Supplementary Text

S1. Archaeology of the Sunghir site

S2. Sequencing, data quality and authentication

S3. Radiocarbon dating

S4. Dataset processing

S5. Mitochondrial and Y chromosome analyses

S6. Relatedness analyses

S7. Identity-by-descent analysis

S8. Population relationships

S9. Admixture graph analysis

S10. Population-specific drift estimation

S11. Demographic inferences

S1. Archaeology of the Sunghir site The Upper Palaeolithic (UP) archaeological site of Sunghir (Сунгирь; Sunghir’;

56°10’34’’ N, 40°30’09’’ E), which yielded the ancient human genomes analysed in this study, is located in European Russia (Fig. S1), ca.190 km east-northeast of Moscow, near the town of Vladimir. This open-air site is situated on a hill ~50 m above the left bank of the Klyazma River. The site was originally covered by several metres of loess, since largely removed by quarrying activities (26). Today the town of Vladimir encroaches on the surroundings of the site (3).

The excavations, which began in the 1950s and have continued intermittently since, have taken place in three areas (Area I to III; Fig. S1). Most of the archaeological material, as well as the elaborately interred human skeletal remains from Burial 1 (a single skeleton, Sunghir 1 [SI]) and Burial 2 (two complete skeletons, and the femur of a third individual [SII-SIV]), which are the focus of our genomic study, were recovered from two main concentrations in Excavation Area II (Fig. S1).

The site, and particularly its rich human burials, has been the subject of numerous anatomical, archaeological and chronological studies (3, 5, 6, 26–30). In the sections that follow, we address a key issue related to our analyses: namely, the lines of evidence supporting the contemporaneity of the burials – and hence our premise that the burials can be treated as a related group. We briefly summarize the burial context, stratigraphy, age, human skeletal and other archaeological remains from Sunghir, and their larger context within the Upper Paleolithic period in Europe (for much more extensive treatments, including of other, later skeletal remains found at the site, see (3, 26, 28, 31)).

S1.1. The burial context Three of the four human remains that are the focus of our analysis were interred together

in Burial 2. The remains of the two sub-adults, Sunghir 2 (SII) and Sunghir 3 (SIII) were laid out in the burial pit head-to-head in an extended position, alongside 16 mammoth ivory spears (31). SII is thought to have been 11-13 years old at time of death, while SIII is estimated to have been 9-11 years of age. There is no evidence that the internments took place at different times; rather, the large burial pit was dug, and the bodies and the spears that serve to ‘join’ the two were placed in it simultaneously. There is no reason to doubt this ‘double burial’ was of two individuals who died at or very near the same time.

The third individual in Burial 2 (SIV) was represented only by the diaphysis (mid-section) of a human femur, which had been polished, hollowed-out and filled with ochre, and carefully placed next to SII. That it is a single element (and not an entire skeleton), and evidence from its bone chemistry which indicates “a different geographic origin for that individual and/or a contrasting post-mortem history,” suggests that the SIV femur section might have been an heirloom piece from an individual who died earlier than SII and SIII. Regardless, that individual presumably had not died significantly earlier than SII and SIII, since his remains were still available to them to collect. Hence, it is of similar age, perhaps someone who had died relatively recently or within generational memory of the individuals in the double burial.

The fourth individual, Sunghir 1 (SI), an adult male, was found nearby, and not in the same burial pit as SII-SIV (3). Demonstrating its contemporaneity with the other humans requires additional evidence.

S1.2. Stratigraphy and geological context The stratigraphy of the Sunghir site, seen in the profile for squares 122 to 127 in

Excavation Area III (3, 26) and representative for the other areas of the site, consists of multiple depositional units and erosional surfaces that together form a stratigraphic sequence ca. 4.5 m thick (Fig. S1).

The basal unit of the excavations (Unit 4 in Fig. S1) consists of light yellowish sandy loam or clayey sands, most probably deposited during cool stadial conditions. Burials 1 and 2 were found within this stratum (3), but were most probably dug into it from the overlying Cultural Layer (Unit 3 in Fig. S1), an organic-rich deposit that ranges in thickness from <20 cm to 1 m, interspersed with lenses of grey sandy loam, and which is associated with artifacts and faunal remains. Based on the single amino acid radiocarbon ages from the human remains, and of a mammoth bone in the Cultural Layer (see Section 1.3), this unit dates to 34,600-33,700 cal years BP.

The Cultural Layer was later heavily disturbed by post-depositional cryogenic processes including solifluction, ice-wedging, and soft-sediment deformation (26). Importantly, however, these effects generally did not extend into the Cultural Layer, nor into the underlying Unit 4. Hence, these cryogenic processes did not disturb the context or archaeological integrity of Burials 1 and 2 (which suffered only slight sediment compaction) or other features dug into Unit 3, such as hearths and pits (26).

Overlying the Cultural Layer is a massive, several meter-thick section of yellowish-tan loam and calcareous loess (Unit 2, with Subunits 2a and 2b in Fig. S1). This loess deposit is marked by manganese streaks, iron staining, occasional sandy lenses, and depositional hiatuses/erosional surfaces (e.g. Surface 5 in Fig. S1) (3, 26) and may have been deposited during the Last Glacial Maximum, suggesting in turn that the cryoclastic processes that disturbed the underlying Cultural Layer might have taken place at this time. The upper reaches of Unit 2 show evidence of landscape stability (i.e. the Surface 5 from which ice-wedges started), and in some areas of the site evidence of soil development (shown in (3): Figure 3.2). The stratigraphic section is capped by a relatively thin (<25 cm) organic horizon, the recent top soil (Unit 1 in Fig. S1).

The stratigraphic position of the burials below Unit 3 and dug into Unit 4 suggests that the burials are broadly contemporary, although ‘real-time’ contemporaneity between Burials 1 (SI) and 2 (SII-SIV) cannot be demonstrated based solely on the stratigraphic position.

S1.3. Radiocarbon chronology The age of the Cultural Layer and burials at Sunghir has been a matter of longstanding

debate. There are today 50 radiocarbon dates on different sample materials from the site and principally from the Cultural Layer, including dates on charcoal, sediment and bone dates on both the associated fauna (mammoth, horse, reindeer), and directly on the human remains SI-SIV. These range widely, from ~31,000-14,000 14C kya BP (Table S1.2).

However, that range can be narrowed. New ages for the human fossils SI-SIV, as well as on a mammoth bone from the Cultural Layer, were obtained using single amino acid (hydroxyproline) AMS radiocarbon dating (5, 6), resulting in far more consistent ages ranging between ~30-29 14C kyr BP. Calibrating these ages using the IntCal13 calibration curve (32) and software OxCal4.2 (33) provides an age estimate of 34.6-33.7 kyr cal BP for both the burials and the Cultural Layer, demonstrating their contemporaneity. These age estimates should be

considered as the most reliable ages on bone because the single amino acid pre-treatment avoids contamination from either conservation activities (e.g., use of preservatives, glues and fumigants) or burial environment more effectively than standard bone sample pre-treatment.

The overlapping AMS dates on hydroxyproline do not prove direct contemporaneity, nonetheless the ages are statistically indistinguishable. Viewed another way, they do not allow us to infer that some could be earlier or later than others. Moreover, as they are the same statistical value, they do not therefore contradict the archaeological evidence for the human remains being contemporary. In effect, they show that the lifetimes of these individuals were not separated by time on the order of centuries, but less than that. S1.4. The Burials: Anthropology and material culture

Burial 1, excavated in 1964, contained the nearly complete skeleton of a 35-45-year-old man (Sunghir 1 [SI]) buried in an extended position lying on his back. SI is estimated to have been 170-180 cm tall, wore footwear throughout his life, but consumed a rough diet that led to pronounced dental wear and ante-mortem loss of teeth (3). Other aspects of SI’s skeleton are further indicators of a relatively harsh lifestyle both during childhood, when he may have suffered from nutritional stress and possibly rickets (34), and as an adult as shown by osteoarthritic lesions in his arms and legs.

SI has a peri-mortem lesion to the lower neck (1st thoracic vertebra) consistent with sharp-force trauma, probably caused by a projectile or knife (35). Whether the injury was the outcome of inter-personal violence or accident has been a matter of debate; Trinkaus & Buzhilova (35) favour an interpretation of hunting accident, but this is unknowable.

Although certain aspects of SI’s life history mirror those of the other individuals, the most striking similarities among them – or at least with SII and SIII, are in the material culture. SI is covered in ochre, especially in the upper part of the body, and surrounded by numerous ornaments and grave goods including ivory bands, a stone pendant, stone tools, and pierced Arctic fox canines. Approximately ~3,000 mammoth ivory beads were found forming a pattern over the body, suggesting that they may have been sewn onto his clothes, probably on inner and outer garments (3, 26, 36).

SII, a nearly complete skeleton (missing only the distal portion of the left arm) was likewise covered in ochre, although less than was found over SI, but as with SI, it was also concentrated on the upper body. SII was wearing 8 ivory arm bands, had a large carving of a mammoth next to his left shoulder, and an animal shaped pendant on his chest. Approximately 5,000 mammoth ivory beads were found covering the body, again probably sewn onto his clothes. Additionally, a string of ~250 Arctic fox canines was positioned around his waist, which could have also been sewn on clothing or onto a belt. An ivory pin found on his chest was probably used as a clothes fastener (3). SII had an estimated stature of 155-160 cm.

The SIII skeleton was decorated in the same fashion as SI and SII, with larger ochre concentrations about the head and shoulders, on the pelvis, and a small amount along the side of the right leg. About 5,400 ivory beads were found on and around the skeleton. SIII was wearing 13 ivory bracelets around both upper and lower arms and the wrists. An ivory pin or clothes fastener was found on the chest (3). SIII had an estimated stature of 139-142 cm.

The beads recovered with SI and with SII-III are identical in shape as well as manufacturing process, even if the beads of individual SI are larger than those found with SII-

III. This suggests that those who buried SI-SIII shared the same material culture traditions and hence most probably belonged to the same cultural group.

Both SII and SIII show, as does SI, evidence of developmental stress and growth disruption in the form of enamel hypoplasias and Harris Lines, probably the result of episodes of resource stress (37, 38). In addition, SII and SIII each may have had a congenital condition that affected growth. SII had no dental wear and lacked development of his masticatory muscles (3, 39). He also had osteolytic cysts of unknown aetiology, and a peri-mortem injury (3). SIII shows very pronounced anterior bowing of the femora, thought to be congenital in origin, although the cause is unknown and much-debated (38, 40–42). SIII was nonetheless strong and physically active, and shows hypertrophy of the right arm, perhaps as the result of compensation for problems with his legs (3, 37). S1.5. The case for contemporaneity

In so far as the archaeological evidence ever allows, there can little doubt that SII and SIII (the two sub-adults) were likely alive at same time, and we can more confidently conclude that their deaths were close to one another in time. This was not an instance of different groups using same the burial spot; there is no evidence, at least within Burial 2, of a later, intrusive burial.

The hollowed-out femur of SIV was likely at or near contemporary with SII and SIII in real time, in so far it was interred with them and, as is now known from this study, was highly similar genetically and genomically. Was SIV someone that SII and SIII might have known in life? We cannot answer that with the available evidence, but it is reasonable to infer that this individual may have died sometime around the time that SII and SIII were alive, or perhaps within a generation or two, allowing them to collect and carry within them until their deaths that part of his skeleton.

SI was different in terms of his mtDNA haplotype, but the striking similarity between this individual and SII and SIII in terms of their material culture, their similar geological and stratigraphic context in the site, and their overlapping radiocarbon ages, indicates they were all archaeologically contemporaneous, and part of the same cultural group. This does not prove SI was contemporary with the others in real time, but he was close enough in archaeological time that his relatedness to the others is relevant to the question of exogamy in this closely culturally-related group. S1.6. Other remains from Sunghir

As noted in the main text, we attempted genome sequencing on two additional Sunghir individuals, although for different reasons they were not ultimately a part of this study. Sunghir 5 (SV), an incomplete older adult cranium of undetermined sex, was the first human fossil discovered at the site (3). It was found in the Cultural Layer above Burial 1 ((3): Figure 3.2), although their precise stratigraphic relationship is uncertain, since SV had been moved as a result of solifluction. The remains were found associated with a large flat stone, and, similarly to the other burials, were covered in ochre. An Arctic fox canine and an ivory bead were found in association with the specimen. A new radiocarbon age (26,042 ± 182 14C BP) obtained as part of this study (Table S1; Supplementary Materials S2) shows SV is younger than SI-SIV.

Sunghir 6 (SVI) is a partial mandible of a young female (?) adult of uncertain stratigraphic origin found in the vicinity of Burial 2bis (3). A new radiocarbon age obtained as

part of this study (Table S1; Supplementary Materials S2) shows it to have been part of an intrusive recent burial.

S1.7. The Sunghir burials in the context of the Mid Upper Palaeolithic burial record of Europe

The Sunghir human remains, their burial circumstances and association, and the genetic relationship between the individuals described in this study point to complex mortuary and funerary practices among the culturally and biologically related group of people who used Sunghir for settlement. Below we explore how Sunghir fits in the overall picture of human burials and UP mortuary and funerary rituals.

A total of 26 single and 8 multiple burials are known (43, 44) between the first appearance of modern humans in Europe and the Last Glacial Maximum (regardless of whether modern humans first appeared in Europe 50-43.5 kyr cal BP (45–51) or only after 40 kyr cal BP (52, 53)). Of these the majority show red ochre colouring (74%) and have ornaments in the burial (71%), while 28.6% also have other grave goods (such as spears, bone disks, etc.). Most buried individuals are male, juveniles are relatively frequent, and include a high incidence of pathologies (43).

During the EUP, the early part of this period (50-45 to 35 kyr cal BP), modern human burials in Europe are rare. The only example is the burial of a 20-25-year-old male at Kostenki 14 (Russia), directly dated to ~38.7-36.2 kyr cal BP, with a clear burial pit cut into the deposits below Cultural Layer III at the site. Red ochre has been documented in the burial pit, which included a hare vertebra and scapula, a mammoth phalange, and three lithics. Most other EUP human remains consist of isolated individual remains within settlement contexts (e.g., Kostenki 14-Layer IVb (54)) or without archaeological context (e.g., Peştera cu Oase, Romania (55)), or for which, based on the poor excavation records, no secure archaeological association can be shown (e.g., Mladeč Cave, Czech Republic (56) and Kent’s Cavern, UK (57); but see (58, 59)). At Grotta di Cavallo, Italy, there is ongoing debate regarding the association between human remains, stone tools and dated shells (60–63). From the beginning of the MUP (Gravettian and other technocomplexes, such as the Gorodsovian in Russia) at ~35-34 kyr cal BP, burials with and without grave goods become more frequent, and include individual and multiple burials. Individual burials include, among others, those from the Dolní Věstonice I and II and Pavlov I sites in the Czech Republic (64, 65), Kostenki 15, Kostenki 12, and Kostenki 18 in Russia (66), the Balzi Rossi Caves (67, 68), Arene Candide (‘il Principe’) (69), Grotta Paglicci (44), and Veneri in Italy (44) (Pettitt 2011), Lagar Velho in Portugal (70), Paviland in the UK (44), and Cro-Magnon and Cussac in France (44).

Multiple burials are much rarer. Besides Sunghir, these include the triple burial at Dolní Věstonice II in the Czech Republic (71), the double burial at Grotta dei Fanciulli (68) and the triple burial at Barma Grande in Italy (67), and the double burial at Krems-Wachtberg in Austria (72, 73) (Table S4). A special case is the multiple burial at Předmostí, Czech Republic, with the remains of 20 individuals in a pit of about 4 x 3.5 m in size (74). All the skeletons are incomplete and the majority are young (< 30 years in age). The excavation took place in the 19th century, making a detailed assessment of the taphonomy and potential funerary rituals difficult (74).

Ochre was found in the majority (74%) of MUP burials (43), either all over (e.g., Lagar Velho, Krems-Wachtberg) or concentrated on certain parts of the body (e.g., Sunghir, Dolní Věstonice II). In some cases, the ochre is concentrated on the head (e.g., triple burial of Dolní

Věstonice II), which is often also the main concentration of beads/pendants (e.g., triple burial of Dolní Věstonice II, Arene Candide (‘Il Principe’), Grotta Paglicci, Grotta del Caviglione, Barma Grande 5, Veneri, and the triple burial of Barma Grande) (44). When in open air sites, the burials occur both in the main settlement areas (e.g., Sunghir (26), Dolní Věstonice II (71)), as well as in more peripheral areas (e.g., Krems-Wachtberg (73)).

Many MUP human remains also occur as isolated individual bones within settlement contexts (e.g., Dolní Věstonice II (65), Grub-Kranawetberg (75, 76)). It is not clear whether these isolated human remains in settlement areas occur because they represent either re-worked/re-deposited burials, or they formed part of the normal waste accumulating in such campsites, or the intentional removal and preservation of body parts of deceased group or non-group members within the settlement. The latter hypothesis is supported by the fact that some of these isolated human remains received special attention by MUP humans, as in the case of SIV, the adult femoral diaphysis from the double burial of the juveniles at Sunghir that was polished, hollowed-out, and filled with ochre. In this case, this study shows that this individual femur was a preserved body part from a group member related to the children buried in the grave where it was found. Other examples include perforated human teeth that were used as pendants (e.g. at Abri Pataud (77), and Les Vachons in France (44), at Pavlov I (65), and Dolní Věstonice I (65)). The Pavlov I perforated human tooth (PI-25) was found in association with six pierced animal teeth and three shells, and it has been suggested that it might have been part of a larger composite object (65). These examples are glimpses of mortuary and funeral rituals that are not solely primarily inhumation but might also involve exhumation and circulation of body parts of the deceased.

The burials discovered at Sunghir are extraordinary in the richness of objects within the graves, as well as in the unusual head-to-head position of the two juveniles. However, they are not altogether unique among MUP burials in western Eurasia, among which there are several examples of multiple individuals within a single grave, bodies covered in ochre or associated with multiple ornaments and other artefacts, or including isolated bones and/or teeth from other individuals within the grave. Complex mortuary and funerary rituals, such as those revealed by the burials at Sunghir, were clearly part of MUP cultural practices.

S2. Sequencing, data quality and authentication S2.1. Samples and laboratory methodologies applied S2.1.1. Samples

The samples used in this study were collected at the Institute of Anthropology and Ethnology of the Russian Academy of Science, Moscow, Russia. They consisted of: a fragment of the left femur (SI_389) and the lower right third molar (M3, SI_388) of Sunghir 1; a fragment of the right tibia (SII_384) and the upper right first pre-molar (P3, SII_383) of Sunghir 2; a fragment of the right femur (SIII_387), the upper left first molar (M1, SIII_385)) and the lower left molar (M1, SIII_386) of Sunghir 3; a fragment of the Sunghir 4 femur (SIV_392); both petrosal bones of Sunghir 5, and; a fragment of the Sunghir 6 mandible (SVI_393).

S2.1.2. DNA extraction

DNA extraction and sequencing library construction steps were performed at the Centre for GeoGenetics, Copenhagen, Denmark, in ancient DNA facilities dedicated to the analysis of ancient hominin samples and physically separated from post-PCR and modern DNA laboratories.

Any trace of conservation product was removed from the sample surface by using scalpel blades and 70% ethanol. The surface was afterwards abraded using a drill. For teeth, the cementum was targeted for DNA extraction as recommended in (78). Each extraction session included a mock extraction blank where no bone or tooth powder is added to the extraction reagents.

We powdered 130 to 380 mg of bone or 46 to 220 mg of tooth cementum per extraction session, using either a microdismembrator or low speed drill (Table S6). We first followed the extraction procedure described by (79), which is tailored to ultra-short DNA fragments and performs DNA purification on MinElute (Qiagen) silica-columns. Briefly, we pre-digested the sample in 1mL extraction buffer (0.45M EDTA, 0.25mg/mL proteinase K) for 1h at 37°C. After centrifugation for 2 min at 16,000g, we discarded the supernatant and incubated the pellet over night at 37°C in 1mL of fresh extraction buffer. After centrifugation for 2 min at 13,000 rpm, we kept the remaining pellet, if any, for further digestion in 1mL of fresh extraction buffer at 37°C for 24 to 48h. This step was repeated if necessary for a third digestion round in 1mL of fresh extraction buffer at 37°C for 72 to 96h. In cases when more than 24h of digestion was performed, 0.25mg of fresh proteinase K was added after each 24h of incubation.

After centrifugation for 2 min at 13,000 rpm, we transferred the supernatant mixed with 13 mL of binding buffer (5M guanidine hydrochloride, 40% isopropanol, 0.05% Tween 20, 90 mM sodium acetate) on a Zymo-Spin V reservoir (Zymo Reasearch) fitted on a MinElute column (Qiagen). The reservoir-MinElute device was centrifuged at 3,000 rpm for a total of up to 8 min. The MinElute column was then placed on a collection tube and centrifuged at 6,000 g for 1 min. The column was washed twice with 750 µl PE and dry-spin at 13,000 rpm. The DNA was eluted twice in 28 or 33 µl EB, after 15 min incubation at 37°C, and all extracts were stored at -20°C in siliconized tubes.

For the Sunghir 5 petrosal bones, we used slightly different extraction methods: a silica columns-based method adapted from (80) and described in (81) for the right petrosal bone; and a silica pellets-based method optimized in (82) for the left one.

A fraction of each DNA extract was directly built into Illumina sequencing libraries (83, 84) to validate the presence of signatures of post-mortem DNA damage. The rest of the DNA extract was subjected to Uracil-Specific Excision Reagent treatment (USER, NEB), to limit the impact of nucleotide mis-incorporations in the downstream analysis (85).

Each of the following steps (USER treatment, library building and library PCR enrichment) was accompanied by a negative control where the sample is replaced by EB buffer (Qiagen).

S2.1.3. USER treatment

For a fraction of each DNA extract, we reduced the importance of the most frequent post-mortem DNA damage, namely cytosines that have been deaminated into uracils, using USER treatment (consisting in a mixture of Uracil DNA glycosylase and DNA glycosylase-Lyase Endonuclease VIII), as proposed in (86). More specifically, a volume of 32.5 µl of the ancient DNA extract was incubated with 10 µl of USER enzyme mix (1U/ µl, NEB reference M5505) for 3h at 37°C. If not processed immediately for sequencing library building, the repaired DNA solution was stored at -20°C.

S2.1.4. Library building, PCR enrichment and sequencing

Illumina sequencing libraries were built following a protocol described in (84) and based on an end-repair step of 21.25 µl of the DNA solution, a ligation step with 12.5 pmol blunt-ended adapters (83) and an adapter fill-in step. We used the NEBNext DNA Library Prep Master Mix Set for 454 (NEB reference E6070) without the ssDNA isolation module. Libraries were enriched and indexed by performing 6 to 14 PCR cycles in 50 µl reactions using 5U of the AmpliTaq Gold DNA polymerase (Life Technologies) as described in (84). After purification on MinElute columns and elution in 20 µl EB following 15 min incubation at 37°C, library concentration and size were checked on a Bioanalyzer 2100 High-Sensitivity Assay (Agilent). Whenever concentrations were not directly compatible with Illumina sequencing, the purified PCR product was re-amplified in four parallel 50 µl reactions for 6 or 7 additional cycles, before being pooled, purified and checked on the Bioanalyzer again.

Final libraries were pooled and Paired-End sequenced (75 to 150 bp read length) on a HiSeq 2500 (Illumina) at the Danish National High-Throughput DNA Sequencing Centre, University of Copenhagen, Denmark.

S2.2. Read processing and mapping

Illumina reads were processed and aligned to the human reference genome hg19 available from the UCSC genome browser

(http://hgdownload.cse.ucsc.edu/goldenPath/hg19/chromosomes/) using the PALEOMIX pipeline (87) with default parameters, except for the minimal mapping quality threshold, which was set to 30, and seeding, which was disabled. PALEOMIX implements in an automatic manner all the analyses described in (88) and (89), which enabled previous characterizations of ancient and modern genomes. Briefly, Illumina sequencing reads were trimmed for adapter sequences using AdapterRemoval (90). When overlapping for at least 11 nucleotides and with a maximal edit distance of 1, paired-end reads were collapsed and base quality-scores at overlapping positions were recalculated using the quality scores of both reads at each overlapping position (87). Collapsed reads were further treated as single-end reads. Trimmed reads and collapsed reads were aligned against hg19 using bwa version 0.5.9-r26-dev (91), filtered for duplicates and re-aligned around indels using GATK (92). Uncollapsed read pairs were filtered from the final BAM alignment file, as this fraction corresponds to relatively long templates and is therefore likely to be enriched in contaminating templates of modern origin. We repeated the full procedure to generate read alignments against the revised Cambridge reference mitochondrial sequence (rCRS, Accession Number NC_012920). Summary statistics, including

the total number of Read Pairs generated, endogenous DNA content per library (the fraction of high-quality hits mapping uniquely against the reference genomes considered), and the contribution of each library to the final coverage estimate, were directly obtained from PALEOMIX (Table S6). For all Sunghir 5 libraries sequenced, endogenous DNA content was lower than 0,1%, therefore this individual was excluded from further processing.

The existence of nucleotide mis-incorporation and DNA fragmentation patterns, which represent the signature of post-mortem degradation reactions (85) was investigated using alignment BAM files and mapDamage2 (93). DNA libraries that were not USER treated showed a progressive decline of CàT substitution rates from read starts, mirrored by a progressive increase of GàA substitution rates towards read ends (Fig. S3A). Additionally, genomic positions preceding read start coordinates were found to be enriched in purines, particularly in Guanine residues. Genomic positions following read end coordinates were found to be enriched in pyrimidines, particularly in Cytosine residues. This is in agreement with previous reports, suggesting depurination as the main driver for post-mortem DNA degradation. Read alignments from USER-treated DNA libraries showed CàT and GàA misincorporations restricted mostly to read-starts and read-ends (Fig. S3B), in line with previous observations suggesting low base excision performance for bases located at the very ends of ancient DNA molecules (13). The base composition of the genomic coordinate preceding read-starts was enriched in cytosine residues, as expected from the enzymatic fragmentation performed, which takes place 3’ of uracil residues. Overall, this indicates that USER treatment significantly reduced the level of nucleotide-misincorporations resulting from post-mortem DNA decay.

S2.3. Sex determination

Sex determination for all individuals was carried out by analysing the fraction of high quality reads (MAPQ ≥ 30) mapping to the Y chromosome (94). All five Sunghir individuals were unequivocally determined as males (Table S7)

S2.4. Contamination

As all individuals were determined to be males, we estimated levels of contamination from present-day humans on the X chromosome using a maximum-likelihood framework implemented in ANGSD (95) (version 0.902). For each individual, we ran ANGSD using the following filters

Exclude pseudoautosomal regions Mapping quality ≥ 30 Base quality ≥ 20 The levels of contamination reported in Table S5 correspond to the maximum-likelihood

estimator using the unbiased sampling-based approach (96) implemented in ANGSD (“Method 2”)

S2.5. Error rates

We obtained error rates for all ancient genomes by estimating rates of excess derived alleles compared to a high quality genome (NA12778, 1000 Genomes (97)) in each substitution class, as implemented in ANGSD (95). We find that error rates are low (< 0.2%) across all individuals and substitution types (Fig. S4).

S3. Radiocarbon dating S3.1. Method descriptions

The two samples from Sunghir 5 and 6 were dated as described below. Conventional radiocarbon ages were obtained in years BP after (98) S3.1.1. Bulk collagen dating

For Sunghir 6, the sample was prepared using the routine Oxford ultrafiltration of collagen method (coded AF*). 750 mg of bone powder was first treated with a sequence of solvents; acetone (50°C, 30 minutes); methanol (50°C, 30 min) and chloroform (RT, 30 min). This sequence is designed to remove potential contaminants such as additives, adhesives, glues or consolidants observed on the surface of the specimen when the sample was drilled. The collagen was then extracted using the standard protocols (99). This comprised sequential washes of 0.5 M HCl to decalcify the bone, 0.1 M NaOH (30 mins) to remove humics, followed by 0.5 M HCl (15 min) at RT. Ultrapure MilliQ™ water rinses were given in between each stage. ‘Collagen’ was gelatinized in pH3 solution at 75°C for 20 hr after which it was filtered with a pre-cleaned Ezee™-filter (Elkay, UK). The filtrate was pipetted into a Vivaspin 15™ 30 kD MWCO ultrafilter and centrifuged until 0.5-1.0 mL of the >30 kD fraction remained. This was freeze-dried and weighed. The ultrafiltered collagen was combusted in an elemental analyser and the CO2 graphitised prior to AMS dating. The C/N atomic weight ratio and carbon and nitrogen stable isotopic values of the collagen were measured using a Sercon mass spectrometer against alanine standards. The calibration of the radiocarbon results was undertaken using the INTCAL13 calibration curve (100) and the OxCal 4.2 programme (33). S3.1.2. Amino acid dating

For the bone from Sunghir 5 (P40,815), a total of 1330 mg of bone were treated. The initial acid-base-acid steps outlined above were applied to extract unpurified ‘collagen’. A total of 16.4 mg of ‘collagen’ was obtained. For the sample from Sunghir 6 (P38722), 50.01 mg of the collagen obtained using the AF* procedure above was taken. Both collagen samples were hydrolysed using 6M hydrochloric acid and then evaporated to dryness in a Genevac EZ-2 vacuum evaporator (Genevac Ltd, Ipswich, UK). 900µl of MilliQ™ deionised water was added to re-dissolve the sample. This was then loaded into a 2 mL BP Plastipak™ syringe fitted with a Thermo Scientific 0.2 µm PTFE syringe filter to remove any insoluble matter and filtered into a Waters® HPLC 1 mL total recovery vial (Agilent Technologies). 200 µL of MilliQ™ water was added to the amino acid residue and filtered into the same HPLC vial to recover as much sample as possible.

Chromatography experiments were performed on a Varian ProStar HPLC system equipped with an autosampler (Model 410), two isocratic pumps with titanium heads (Model 210), a column oven set at 30°C containing a Primesep A preparative column (22 × 250 mm, particle size 5 µm; SIELC, IL, USA), a UV detector (Model 320) set at 205 nm and a fraction collector (Model 701). The system is controlled by Star workstation PC software (Version 6.0). The autosampler was modified with a 1 mL glass syringe and a 2 mL sample stainless steel loop, enabling up to 1 mL of sample to be injected. The separation was achieved using a gradient of MilliQ™ deionised water (eluent A) and 0.3% phosphoric acid diluted with MilliQ™ deionised water (eluent B), as described in Table S8 and at a total flow rate of 18 mL/min.

For the sample from Sunghir 6, the collected water fraction containing hydroxyproline was concentrated using a Genevac EZ-2 Plus vacuum evaporator until totally dried. For the sample from Sunghir 5, because the sample was too small for the dating of only the hydroxyproline, glycine, proline and alanine were also collected and concentrated using the same procedure. The amino acid(s) was/were then reconstituted in 25 µL of MilliQ™ deionised water and loaded on to 12 mg of Chromosorb™ in cleaned tin capsules. Stable isotopic measurement for carbon and nitrogen, combustion and graphitisation were performed as described in (99) S3.2. Radicarbon dating results S3.2.1. Sunghir 6

This sample was first dated using the ORAU routine ultrafilter (AF*) procedure and produced a date of 884 ± 23 BP (Table S9, Fig. S3). The only doubt regarding the reliability of the result concerned the possibility that the same contaminants used in the consolidation of the other Palaeolithic bones at the site were applied to the specimen. The same sample was then re-dated using the “single compound approach” with the isolation and dating of the amino acid hydroxyproline (Table S6). The results are statistically identical, suggesting there is no significant contamination in the bone that was taken for AMS dating. These data confirm that this sample is not associated with the UP burials at Sunghir.

S3.2.2. Sunghir 5

The sample from Sunghir 5 was dated using the “amino acid approach”. Because of the poor preservation of the collagen in this sample, it was not possible to date only the hydroxyproline. Four amino acids were collected, namely Hydroxyproline, Glycine, Proline and Alanine. They were then recombined and dated by AMS (Table S10, Fig. S4). We obtained a date of 25240 ± 160 BP and, after correcting the measurement for the HPLC background after (101), we calculated an age of 26042 ± 182 BP.

S4. Dataset processing S4.1. Overview of dataset processing

Datasets for population genetics analyses of were constructed by combining the Sunghir individuals and other previously published ancient genomes with different panels of modern reference populations. In order to minimize batch effects between datasets from different sources, genotypes for all ancient individuals were obtained from BAM files using a common pipeline. Individuals with low coverage shotgun data or from targeted capture were genotyped by picking the majority allele for the genomic positions in the respective panel, restricting to reads with mapping quality ≥ 30 and base quality ≥ 30. If both alleles were present at a site with the same coverage, we selected a random allele. For all higher coverage individuals (≥ 9X; Table S11), per-individual diploid genotypes were called using samtools mpileup (-C50 option; version 1.1) and bcftools call with the consensus caller (-c option; version 1.2-4) (102, 103). Calls from each genome were filtered for a minimum depth of 6 reads or 1/3 of the average sequencing depth if this was higher than 6. Additionally, the calls were filtered for a maximum of 2 times the average depth to rule out possible copy number variation. Variants were subsequently filtered out if there were two variants called within 5 base pairs of each other, Phred posterior probability of less than 30 or strand bias or end distance bias p-value < 104. Per individual calls were merged across all samples using GATK CombineVariants (version 3.3-0) and filtered for deviations from Hardy-Weinberg Equilibrium with p-value < 104. For whole genome reference sets, analyses were furthermore restricted to regions that were both within the 1000 Genomes Phase 3 strict accessible genome mask (ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/supporting/accessible_genome_masks/20141020.strict_mask.whole_genome.bed), as well as outside repeat regions (UCSC genome browser simpleRepeat table).

S4.2. 1240K capture panel

This dataset is based on a set of SNPs targeted in two panels of in-solution capture used in previous ancient DNA studies (104, 105), which included a total of 1,153,048 autosomal SNPs. We extracted genotypes for our newly reported genomes as well as a large panel of previously published ancient individuals at those SNP positions, requiring that the genotyped alleles match either of the two alleles targeted with the capture probes. In addition to the high coverage individuals described above, we included individuals from the following sources:

- Targeted capture and whole-genome shotgun data from Paleolithic and Mesolithic

Eurasian HGs (105–107, 11, 81, 108, 109, 8) - Targeted capture and whole-genome shotgun data from Early Neolithic farmers from

the Middle East and Europe (109–114) - Whole-genome shotgun data from three modern Mbuti Pygmy individuals to use as an

outgroup (115)

S4.3. Whole-genome sequencing panel To obtain a diverse panel of whole-genome sequencing data we initially merged all high

coverage ancient genomes (Table S11) with data from a number of recent datasets of modern high coverage genomes:

- 256 individuals from world-wide populations (115) (the fully public ‘C-Team' SGDP genomes)

- 25 individuals from world-wide populations (7, 13) (“A-Team” and “B-Team” genomes)

- 83 Aboriginal Australian individuals (116) - 10 Andamanese Islanders (117) - 26 individuals with Siberian and Native American ancestry (118, 119) With the exception of the 256 SGDP genomes, all individuals were genotyped individually

from BAM files using the same pipeline as the high coverage ancient individuals. For the SGDP, we downloaded already genotyped emit-all VCF files and merged them with the data from the other individuals (with filters as described) using ‘bcftools merge’(103), requiring a read depth ≥ 6 across all SGDP genotypes. The final merge was then obtained by restricting to bi-allelic SNPs within the “accessible genome” region masks (see above). Genotypes for low-coverage ancient individuals were extracted for this final set of SNPs as described, requiring that the sampled allele matches either of the two alleles in the high-coverage panel. We added ancient individuals from the following sources:

- Whole-genome shotgun data from Paleolithic and Mesolithic Eurasian HGs (8, 81, 106–

108, 120) - Whole-genome shotgun data from Early Neolithic farmers from the Middle East and

Europe (110, 112–114) - Whole-genome shotgun data from Neanderthal individuals (7, 121)

S5. Mitochondrial and Y chromosome analyses S5.1. Mitochondrial DNA

To reconstruct mitochondrial DNA sequences for the study individuals, we called haploid genotypes for reads mapped (MAPQ ≥ 30) to the Cambridge reference sequence (rCRS; Genbank NC_012920.1) using endoCaller from schmutzi(122), and estimates of contamination were performed using mtCont from the same package. Genotypes with a posterior probability exceeding 50 on a PHRED scale (probability of error: 1/100000) were retained. Haplogroups were then inferred from variant positions using HaploGrep(123). Multiple sequence alignments were performed using prank v.150803 (124) and used to build a maximum-likelihood tree with RAxML(125) using the GTRGAMMA model (Fig. S7).

S5.2. Y chromosome S5.2.1. Haplogroup assignment

We called haploid genotypes for reads mapped (MAPQ ≥ 30) to the Y chromosome using samtools/bcftools (102, 103). To determine haplogroups, we extracted genotypes for all individuals at SNPs included in the Y-DNA haplogroup tree from the International Society of Genetic Genealogy (ISOGG, http://www.isogg.org, version 10.107). All Sunghir individuals clustered with Haplogroup C1a2 (Table S12-S15), which is rare in contemporary Europeans. Related haplogroups were also previously reported for other early Eurasian HGs, including Kostenki 14 (8, 11) and a 7,000 year old Mesolithic HG from La Braña, Spain (126).

S5.2.2. Phylogenetic analysis of Sunghir 3

To further investigate the position of the Sunghir individuals in the phylogeny of contemporary Y chromosomes, we analysed the sequence of the higher coverage individual SIV together with a subset of 58 individuals covering haplogroups A, B, C, D and G from a recent Y chromosome dataset (127). We restricted this analysis to the regions passing all filters in (127), and that had coverage in SIII. Phylogenetic tree reconstruction was carried out using the maximum likelihood algorithm implemented in MEGA (128) using the Tamura-Nei model and 100 bootstrap replicates. Sunghir 3 clusters with an individual from Nepal (nep-0172; 96/100 replicates) carrying the C1a2-defining V20 mutation, albeit with an early divergence close to the split with haplogroup C1a1 (represented by individual JPT-NA18974 from Japan) (Fig. S8). The deep divergences and widespread geographical distribution observed in the descendants of these haplogroups suggest a rapid dispersal of these lineages during the Upper Palaeolithic.

S6. Relatedness analyses We sought to determine the familial relationship between the Sunghir individuals by

inferring rates of identity-by-descent (IBD) sharing between pairs of individuals. As not all study individuals are of sufficient genomic coverage to accurately call diploid genotypes, we aimed to infer their relationships directly from genotype likelihoods, using three different approaches:

- Estimation of global IBD sharing coefficients (k0, k1, k2) using the maximum-likelihood

approach implemented in ngsRelate (129). - A method inferring local IBD sharing using a hidden Markov model, implemented in

relate (130) - The kinship coefficient estimator implemented in KING (131), based on pairwise

identity-by-state (IBS) sharing. The IBS sharing matrix between two individuals was obtained by estimating the 2D-SFS for each pair using realSFS from the ANGSD package (95)

Expected values for the different estimators and selected degrees of relatedness are shown

in Table S16. A drawback of methods (1) and (2) is that they require estimates of population allele

frequencies from unrelated individuals at the SNPs used in the respective analysis. As we do not have good estimates of these frequencies for the Sunghir populations, we evaluated two different alternative approaches:

- 1000 Genomes sites and European (CEU) allele frequencies, as Europeans are the

closest modern genetic relatives to our individuals - Allele frequencies estimated from the study individuals combined with 35 previously

published Mesolithic and Palaeolithic individuals (126, 106, 15, 132, 108, 8), restricted to the set of 1240k capture sites

For all analyses, we used genotype likelihoods obtained using the SAMtools genotype likelihood model in ANGSD, restricting to reads with mapping quality ≥ 30 and nucleotides with base quality ≥ 20. We furthermore considered likelihoods only for the two alleles segregating in the respective reference SNP sets. Below we discuss results and their implications for each method.

S6.1. ngsRelate analysis

To assess the robustness of the estimates in the presence of ancient DNA deamination we carried out the analysis both with and without transitions. We found that ngsRelate substantially overestimates both inbreeding and relatedness coefficients when using the 1000 Genomes CEU allele frequencies (Table S17, S18), therefore all following results are based on Mesolithic/Paleolithic frequencies. The model in ngsRelate assumes no inbreeding within individuals, we therefore first estimated inbreeding coefficients for each sample. As inbreeding is negligible in all individuals (Table S19), we proceeded with the relatedness inference. We find a non-zero IBD0 probability (k0) for all pairwise comparisons within the Sunghir sites, which would suggest some amount of relatedness between all individuals (Table S20). The closest

related pairs are Sunghir 3 / Sunghir 2 (the double burial), with k0 values consistent with a third-degree relationship. The results are also consistent whether we use the full set of SNPs or transversions only, however we were concerned that the lack of accurate population allele frequencies could bias these estimates and therefore lead to false positives. To investigate this possibility, we additionally inferred relatedness between individuals from different sites which cannot plausibly be close genetic relatives, but show similar genetic ancestry. We find that while no relatedness between the Sunghir individuals and Kostenki 14 is inferred as expected, later Palaeolithic and Mesolithic HGs with substantial geographic and temporal separation nevertheless show relatedness comparable or even higher than that within the Sunghir sites (Table S20). In particular, the individuals Loschbour and Bichon would be inferred to be 3rd degree related, despite being temporally separated by ~5,000 years. We therefore conclude that the ngsRelate estimates are not reliable, likely as a consequence of the lack of accurate population allele frequencies and due to the fact that ngsRelate does not take into account LD, which often causes relationships to be overestimated (133).

S6.2. Relate analysis

We next explored whether an approach based on local IBD sharing (Relate(130)) could mitigate the effects of imperfect population allele frequencies. For this analysis, we required a minimum of 2 reads for the genotype likelihoods calculation. We furthermore excluded SNPs with a MAF below 5%, and pruned the data for linkage disequilibrium (LD) by removing site with an r2 > 0.2, using PLINK(134). This procedure was carried out independently each pair of individuals and all sites where genotype likelihoods were available for both individuals, to maximize the number of SNPs in each comparison. LD was estimated based on the genotypes from the 1000 Genomes CEU panel (135). Results for this analysis are shown in Tables S21 (CEU allele frequencies) and S22 (Palaeolithic/Mesolithic allele frequencies). We find that relatedness estimated with CEU frequencies shows strong inflation, with close relatedness (2nd degree) inferred for all cross-site comparisons of the Sunghir individuals with Kostenki 14. Results improve substantially when using the Palaeolithic/Mesolithic allele frequencies, after which only a small number of within-site comparisons show relatedness. The closest relatedness was again found for the double burial pair Sunghir 2 / Sunghir 3, which is consistent with a 3rd degree relationship, but we again observe some residual relatedness between the Mesolithic individuals (Table S22).

S6.3. IBS analysis

Given the strong influence of the allele frequencies used in both ngsRelate and relate analyses, we sought to obtain additional relatedness estimates using a method that does not depend on allele frequencies. In particular, we used the estimator introduced by (131), implemented in the KING package (equation (9) i n (131)). We calculated the estimate from a matrix of IBS sharing between two individuals, obtained by estimating the 2D-SFS at 1000 Genomes SNP sites, for each pair using realSFS from the ANGSD package (95). As the KING estimator cannot distinguish parent-offspring and full sibling relationship, we additionally used a related statistic called R1, the ratio of double heterozygous (Aa/Aa) sites divided by the total number of discordant genotypes. Using the notation of (131):

𝑅1 =𝑁%&,%&

𝑁%%,%& + 𝑁%%,&& + 𝑁%&,%% + 𝑁%&,&& + 𝑁&&,%% + 𝑁&&,%&

Results for this analysis are shown in Fig. 1A, where we find that none of the individual

pairs is closely related (third degree or lower). The highest kinship coefficient was again found for Sunghir 2 / Sunghir 3 (0.004), but it nevertheless remained below the threshold for third degree (0.0442; Table S16).

In order to verify that our approach can detect close relationships from low-coverage

ancient DNA data, we repeated the analysis on the individuals captured at 1240K SNP sites in (109), which also contains close relative pairs. Restricting to individuals with ≥ 1X coverage (the lowest coverage in our dataset, Sunghir 1), we obtained a final dataset of 75 individuals, to which we added our six study individuals. Results of this analysis are shown in Fig. 1B and Table S23. Among all pairwise comparisons, we find only four clear outliers with close genetic relatedness. The two pairs inferred as first degree relatives (I0736/I0854; I0114/I0117) correspond to all pairs with coverage ≥ 1X that had been described as 1st degree relatives in (109). The two individuals of the second-degree relative pair (I0744/I1097) are from Anatolian Neolithic site in Barcin, Turkey, and share the same Y chromosome haplogroup. This pair has also been identified as second degree relatives in a recent preprint on relatedness estimation from low-coverage ancient DNA with population-level data(136). We also identify a previously unreported third-degree relationship for one pair of Scandinavian hunter-gather individuals Motala, Sweden. For the Sunghir individuals we again do not find close relationships, consistent with the results from the 1000 Genomes SNP set (Table S23).

Based on these results, we conclude that relationship inference using population allele

frequencies leads to spurious relatedness if accurate frequencies cannot be obtained from the population of the individuals of interest. The IBS-based estimator on the other hand only relies on the pairwise 2D-SFS between the individuals, and appears robust for genomic coverages ≥ 1X. We therefore conclude that the individuals at Sunghir were not closely related (third degree or lower), a result which is additionally supported by the IBD analyses in Supplementary Materials S7.

S7. Identity-by-descent analysis S7.1. Overview of the analysis and datasets

To gain more detailed insights into patterns of genetic relatedness and recent co-ancestry among the UP individuals, we inferred genomic segments identical-by-descent (IBD) and homozygous-by-descent (HBD) in the high coverage whole-genome panel (Supplementary Information 4). We additionally included the intermediate coverage individuals (≥ 4X) from Sunghir (Sunghir 2, Sunghir 4) in this analysis. In order to investigate demographic scenarios compatible with the observed distributions of HBD and IBD tracts, we also performed large-scale coalescent simulations of whole-genome data, using the package msprime (12). We simulated a simple two-population demography that mirrors the divergence time of Sunghir from the European ancestral population (Fig. S9), varying effective population sizes for the “Test” population from 20 to 10,000 individuals. To match the inference on the real data as closely as possible, we simulated whole genomes for 100 haploid individuals in each population, using the actual chromosome lengths and recombination rate maps (HapMap) (137) from the human genome. Haploid individuals were then randomly paired into diploid individuals and written as VCF files for IBD inference, using the output functions implemented in msprime. We assumed a mutation rate of 1.25 x 10-8 / bp / generations (138) for all simulations. S7.2. IBD inference

We inferred genomic segments homozygous-by-descent (HBD) and identical-by-descent (IBD) from the real and simulated datasets using IBDseq (139), separately for three types of datasets:

- All modern and ancient genomes in the high-coverage whole-genome panel, except the

archaic humans - All genomes for each simulated dataset - Archaic humans (Altai Neanderthal, Denisova) combined with five Ju’hoan individuals. We ran IBDseq with default parameters, except allowing for a higher genotype error rate

due to the lower quality ancient DNA data in the callset (errormax = 0.005). For the simulated data, the error rate was set to 0. We furthermore considered only segments with LOD ≥ 2 for all analyses. The archaic humans were treated as a separate dataset, as running them with the full modern dataset resulted in both individuals showing very long tracts. For this analysis, we also used a larger window size and lower cutoff for the LD thinning (r2max = 0.1; r2window = 3000), which resulted in tracts consistent with previous results for these individuals (Fig. S10). IBDseq assumes that the genotype data derives from a homogenous population, since it estimates population allele frequencies from the data. A violation of this assumption, as is the case in our analysis, can lead to spurious short IBD segments. We therefore follow the recommendation of (139) and restrict the analysis to long tracts (≥ 2 Mb HBD; ≥ 4 Mb IBD). Finally, all inferred tracts were then further processed to remove artificially long tracts spanning assembly gaps or centromeres. In particular, tracts that spanned centromeres or with tract length ≤ 1Mb on either side of the gap were split into separate tracts at the respective gap boundaries. Genomic locations for assembly gaps were obtained from UCSC genome browser (http://hgdownload.cse.ucsc.edu/goldenPath/hg19/database/gap.txt.gz).

S7.3. HBD / IBD tracts validation

We were concerned that increased error rate due to lower coverage and ancient DNA damage in some individuals could lead to artificially break up long HBD and IBD tracts, as errors would increase the apparent heterozygosity across the genome. While IBDseq allows for sequencing errors in the inference, we also manually inspected the distribution of genotypes along the genome for Sunghir individuals. We find that inferred HBD tracts correspond well to regions of low heterozygosity, suggesting that effects of tract shortening are negligible. An example for genotypes and HBD tracts on chromosome 18 is shown in Fig. S11.

We furthermore compared HBD tracts obtained using IBDseq to those obtained applying the sliding window approach of (7) for SIII, Ust’Ishim as well as archaic humans Altai Neanderthal and Denisovans. Tracts inferred for modern humans are highly concordant between methods for SIII (Figure S12), whereas some differences are observed for archaic humans, particularly for shorter tracts (Figures S12, S13), likely due to poor allele frequency estimates for those individuals. Furthermore, the sliding window approach tends to overestimate lengths in Ust’Ishim, leading to an apparent excess of HBD tracts called by this approach (Figs. S12, S13). Nevertheless, the sliding window also approach identifies a large excess of very long HBD tracts in the Altai Neandertal compared to the UP modern humans, consistent with the IBDseq results (Figure S14).

S7.4. Relatedness from IBD tracts

We used the inferred IBD tracts between Sunghir individuals as an additional test for relatedness, as closely related individuals are expected to share long genomic tracts IBD. As positive controls, we included known close relatives from Aboriginal Australians (116) as well as Andaman Islanders (117) in the whole-genome panel in this analysis (Fig. 2B). We find that individuals from Sunghir generally share less total IBD in shorter tracts than known second degree relative pairs. These results therefore suggest that our individuals are not closely related, consistent with the relatedness analyses (Supplementary Materials S6).

As sequencing errors can lead to the breaking up of longer tracts and therefore possible underestimation of genetic relatedness, we further investigated IBD sharing between the closest pair (Sunghir 2 / Sunghir 3) using an alternative approach not relying on genotype calls. We divided the genome in non-overlapping 100kb windows, and estimated the 2D-SFS for the individual pair within each window directly from genotype likelihoods, using realSFS(95). As regions of IBD sharing are not expected to contain discordant homozygote genotypes (IBS0), we investigated the distribution of the counts of those bins in the SFS along the genome. Results for this analysis are shown in Fig. S14. We find that IBD tracts inferred from genotypes using IBDseq correspond well to extended regions of low IBS0 counts, interspersed with high IBS0 windows. We therefore conclude that IBD tracts called using IBDseq are sufficiently accurate to rule out long IBD sharing and close relatedness between Sunghir 2 / Sunghir 3.

S7.5. Effective population size from IBD tracts

We estimated NE from both the inferred IBD and HBD tracts using the method described in (140) (equation 20), restricting to segments ≥ 4 Mb.

S8. Population relationships S8.1. Overview of the analysis and datasets

We carried out different analyses to investigate the broader genetic affinities of the Sunghir individuals. All population genetic analyses were carried out using the 1240K capture panel (Supplementary Information 4), as many of the previously published individuals in this analysis had their data generated only by in-solution capture of this targeted SNP panel. We restricted this analysis to ancient individuals with ≥ 30,000 SNPs covered in the 1240K panel. Unless otherwise noted, results for Kostenki 14 are based on the whole-genome shotgun genome reported in (11). f- and D-statistics were calculated using the allele-frequency based estimators as described in (141), using allele frequencies calculated from genotype calls. For analyses of groups with multiple ancient individuals, genotypes for low coverage individuals were counted as haploid alleles for allele frequency estimation.

S8.2. Principal component analysis

As a first assessment of the genetic affinities of the study individuals we carried out principal component analysis (PCA), as previously described (82). In particular, we projected the low coverage ancient individuals onto the PCs inferred from different sets of modern and high coverage ancient individuals, using the ‘lsqproject’ option in smartpca from the EIGENSOFT package (142). Results for reference panels of all non-African populations, as well as restricted to West Eurasian populations are shown in Fig. S16 and S17.

In both reference panels, the Sunghir individuals form a tight cluster close to the two

individuals from Kostenki (Kostenki 12; Kostenki 14). All individuals are shifted towards West Eurasia on the East-West cline represented by PC1 in the non-African PCA. When restricting to present-day reference individuals from West Eurasia, Sunghir and other UP HGs fall close to the origin, with a subtle shift towards Northern European populations, reflecting shared ancestry with later West Eurasian HGs and the North-South gradient of this ancestry in modern Europeans previously described (8, 11, 120).

S8.2. Outgroup f3 statistics

Genetic clustering of ancient individuals only using outgroup f3 statistics was carried out using the full 1240K panel SNP set, with three Mbuti individuals from whole-genome sequencing panel (115) as outgroup. For the MDS analysis, we converted the pairwise f3 matrix into a distance matrix by normalizing the values to the range [0,1], and tsubtracting the resulting value from one. MDS was carried out in R (143) using the ‘cmdscale’ function. Results are shown in Fig. 3B, Fig. S18 and Fig. S19, as well as in Additional Data Table S1.

We note the following general patterns: - Individuals from Sunghir share more drift with each other than with any other group. - Among individuals from other groups, Sunghir shares most drift with Kostenki 12 - Other groups with high affinity are Kostenki 14 and individuals of the Vestonice cluster - Both Kostenki individuals share most drift with Sunghir

S8.3. f4 statistics

We calculated f4 to further investigate the genetic affinities of Sunghir individuals. Results are shown in Fig. S20 and Additional Data Table S2.

We note the following observations: - All Sunghir individuals form a clade with each other at the exclusion of the other test

groups (Fig. S20, panels 2, 3 and 4) - Kostenki 12 is closer to Sunghir individuals than to Kostenki 14 (Fig. S20, panel 1) In Supplementary Materials S9 we further investigate these affinities using admixture graph

frameworks.

S9. Admixture graph analysis S9.1. Overview of the analysis and datasets

The analyses of genetic drift sharing of the Sunghir individuals indicated a close affinity to UP Eurasians from Kostenki as well as the Vêstonice cluster (8, 11). To further investigate these affinities we used the admixture graph framework implemented in qpGraph (141). As we found that all Sunghir individuals are consistent with forming a clade, we generally used the high quality diploid genotype calls of SIII only to represent Sunghir in the different graphs we tested. The data used for Kostenki 14 is the whole-genome shotgun data of (11) unless otherwise noted.

S9.2. Sunghir

To determine possible fits for Sunghir we started the analysis with the best-fit model reported in (8) as a base. We replicate the model with two slight outliers (|Z| > 3), both involving the Villabruna sample (Fig. S21). As this individual has been shown to be more prone to errors due to the non-UDG treatment (8) we accept a poorer fit of this individual and proceed with adding SIII.

Motivated by the f-statistic results, we aimed to fit SIII in different positions along the Kostenki 14 lineage that contributed a majority fraction to the Vêstonice cluster. We find that we can fit SIII without additional mixture events as a descendant of that lineage, with more recent shared ancestry with the Vêstonice cluster than Kostenki 14 (Fig. 3C; Fig. S22). Adding SIII as either a sister group or outgroup to Kostenki 14 results in a trifurcation at their common ancestor, as well as a poor fit for the shared drift with Vêstonice (Fig. S23). Furthermore, we also reject SIII being a sister group to Vêstonice descending from the same mixture event, as this leads to a lack of shared drift of SIII with Kostenki 14 (Fig. S24).

S9.3. Kostenki 12

Our f-statistic results indicated a close relationship of Sunghir with the Kostenki 12 individual. In the admixture graph modelling of (8), no attempt to fit this individual in addition to the base topology reported and used as a scaffold here, due to its low coverage. However, it was shown to be a possible replacement for either Kostenki 14 or Vêstonice 16. We sought to investigate whether the additional SIII individual with distinct ancestry could further constrain the position of Kostenki 12 despite its low coverage. Results and Z-scores for the worst-fitting f-statistics are shown in Fig. S22. Despite the low number of markers available in this analysis (11,305 SNPs), we can reject topologies where Kostenki 12 is closer to either Kostenki 14 (Fig. S25A) or a direct member of the Vêstonice cluster (Fig. S25B). The analysis rather suggests a close relationship of Kostenki 12 with Sunghir, albeit with a lack of resolution to further differentiate their relative positions within the Kostenki / Sunghir lineage. As Kostenki 12 is more recent than Kostenki 14 and closer in age to Sunghir, our results suggest that groups closely related to Sunghir likely replaced the earlier people at Kostenki (represented by Kostenki 14).

S10. Population-specific drift estimation S10.1. Overview of the analysis and datasets

To investigate genetic continuity between early UP HGs and their later counterparts in Eurasia, we estimated levels of population-specific drift (i.e. branch lengths) from pairs of genomes after their divergence using a coalescent-based approach described in (144). The method works by fully parameterizing the ancestral allele frequency spectrum from the observed sample configuration of the two genomes (i.e. the two-dimensional site frequency spectrum (2D-SFS) of two diploid genomes). In order to also be able to include lower coverage individuals such as Kostenki 14, we estimated the 2D-SFS for all pairs from genotype likelihoods using the program realSFS (145) implemented in the ANGSD package (95). We further applied the following filters:

- Only SNPs found polymorphic in African populations from Phase 3 of the 1000

Genomes Project (135) - Only transversion SNPs - Mapping quality ≥ 30 - Base quality ≥ 30 - Restricted to regions within the 1000 Genomes Phase 3 strict accessible genome mask

(ftp://ftp.1000genomes.ebi.ac.uk/vol1/ftp/release/20130502/supporting/accessible_genome_masks/20141020.strict_mask.whole_genome.bed)

- Remove repeat regions (UCSC genome browser simpleRepeat table). Drift parameters were then estimated from the 2D-SFS by maximizing the likelihood using

the ‘nlm’ method implemented in the ‘optimx’ (146) package in R (143). We estimated drift parameters for different ancient genomes with a variety of modern and

ancient reference genomes, obtaining results consistent with population affinities observed in the other analyses (Fig. S26).

S10.2. Results

Below we discuss results for each test individual in greater detail. Ust’-Ishim Branch lengths for Ust’-Ishim were found to be similar for all comparisons, consistent

with this individual being an approximate outgroup to most Eurasians (15). The largest drift is observed with ancient Neolithic and contemporary European individuals, in line with their reported “basal Eurasian” admixture (120). The overall levels of Ust’-Ishim-specific drift are low compared to the other test HGs, suggesting that the individual was genetically close to the divergence from the ancestral Eurasian population.

Sunghir / Kostenki 14 We find that SIII shows substantial population-specific drift with all tested individuals,

except the other individuals from the same site. The lowest estimates outside Sunghir are obtained with Kostenki 14, consistent with results from the ancestry analyses. However, despite their affinity, the results also show substantial amounts of drift specific to Kostenki 14 after its

divergence, therefore rejecting a directly ancestral relationship to Sunghir. Estimates are high for both Sunghir and Kostenki 14 when compared to later European HGs, suggesting that despite their shared early European ancestry, they did not form a direct ancestral group to the later European HGs in our dataset.

Bichon The late Palaeolithic HG individual from Bichon shows substantial drift compared to all

other individuals, with the exception of the Mesolithic HG Loschbour. Bichon has an estimated branch length close to zero after its divergence with the common ancestor with Loschbour, indicating population continuity in Central European HGs between the late UP and Mesolithic (108). The long branch lengths with all other reference populations are consistent with the small population sizes in HG groups suggested previously (107).

Bar8 Drift estimates of the individual Bar8 from the Anatolian Neolithic are consistent with a

directly ancestral relationship to Early European Neolithic individuals from SE Europe (Klei10).

S11. Demographic inferences Statistical framework Likelihood inference based on the Site Frequency Spectrum

The parameters of alternative demographic scenarios were inferred using the site frequency spectrum (SFS) by approximating the likelihood of a given model with coalescent simulations (147, 148). All computations were done with an extension of the fastsimcoal2 simulation software (149). Coalescent simulations are performed under specific parameters values q of a given model to estimate the expected entries of the SFS , and the likelihood is then obtained as

where is the observed (multidimensional) SFS, n is the total number of entries in the SFS given by the product of

, ni is the number of gene copies in the i-th deme, S is the number of polymorphic

sites, L is the length of the observed sequence data, and is the probability of no mutation on

the tree, obtained as assuming a Poisson distribution of mutations occurring at rate , where T is the expected tree length. Note that if the data contains linked SNPs this is actually a composite likelihood estimator. The likelihood function in equation (1) thus accounts for the number of monomorphic sites and all the entries of the SFS. We note that an alternative likelihood function can only focus on the entries of the SFS for polymorphic sites, discarding information on the number of monomorphic sites, i.e.

As described in (149), the likelihood is maximized using a conditional maximization algorithm ECM (150), which is an extension of the EM algorithm where each parameter of the model is maximized in turn, while keeping the other parameters fixed at their last estimated value. The maximization of each parameter is done using Brent’s algorithm (151). We start with initial random parameter values, and perform a series of ECM optimization cycles until estimated values stabilize or until we have reached a predefined number of ECM cycles (50, unless specified otherwise).

We used a strategy where we begin by optimizing the full likelihood (eq. 1) for a given number of cycles (i.e. 10) and then optimize LSFS (eq. 2) for the remaining (40) cycles. This strategy aims at maximizing the fit between the expected and the observed SFS. At the end of the run, a rescaling factor is computed as , where and are the observed and

expected numbers of polymorphic sites, respectively. is obtained as , with being the expected tree length for the maximum-likelihood parameters. The final maximum-likelihood parameters are then rescaled by RF in order to produce a number of polymorphic sites equal to those observed: the effective population sizes N’s and times of events T’s (including the

ˆ ip

1 1{ , , }nX m m -= …

1

( 1)d

ii

n n=

= +Õ0P

0TP e µ-= µ

1

1

ˆ i

nm

SFS ii

L p-

=

µÕ

fullL

exp/obsRF S S= obsS expS

expS ˆexpS Tµ= T̂

1

0 01

ˆPr( | ) (1 ) i

nmL S S

full ii

L X P P pq-

-

=

= µ - Õ

age of samples) are multiplied by RF, whereas migration rates m’s are divided by RF, such that Nm parameters are left unchanged. Admixture rates are also left unchanged.

Demography of ancient modern humans (Sunghir and Ust’-Ishim) Data preparation and processing

We selected, for our analyses, the Sunghir individual with the highest coverage (Sunghir3) (Table S5). In our demographic models, we also included one archaic human (Altai Neanderthal) (7), one ancient modern human sample (Ust’-Ishim) (15), and four present-day modern humans (two Sardinians from Europe and two Han Chinese from East Asia), who had been previously sequenced at high coverage (7, 13).

We then considered only autosomal SNPs found outside genic regions (as defined by Ensembl version 71, April 2013 (152)), and outside CpG islands (as defined on the UCSC platform (153)). We finally removed SNPs not found in regions that passed a set of minimal filters used for the analysis of the high coverage Altai Neanderthal genomes (map35_50%) (7, 13), and for which the coverage for Sunghir3 was lower than 15X.

The ancestral state of the SNPs was inferred using the ancestral hg19 genome provided by the 1000G consortium (97), which was inferred from the alignments of six primate genomes and released in the Ensembl Compara 59 database (154).

From our dataset of autosomal SNPs passing the above filters, we generated the multidimensional SFS for these different datasets with the Arlequin software ver 3.5.2.2. This dataset was generated by considering concatenated autosomal segments (blocks) of 1Mb (not necessarily contiguous along the chromosome) that were sequenced in all individuals (i.e., no missing data). We identified 294 such blocks on the autosomes, with a total of 290,022,578 sites, of which 756,415 and 761,385 were SNPs in the Sunghir and Ust’-Ishim datasets, respectively. For the non-parametric block-bootstrap analysis used to infer parameter confidence intervals, we resampled these whole blocks with replacement to generate 100 sets of 294 blocks. Bootstrap datasets have the same cumulative length of 294 Mb, and have numbers of SNPs very similar to the original data set (between 742,020 and 769,861 SNPs for Sunghir3 dataset, between 746,409 and 774,170 for Ust’-Ishim).

We inferred the demography of Sunghir3 and Ust’-Ishim individuals independently. Hence we analysed two datasets with the following individuals: (i) Sunghir3, Altai Neanderthal, 2 Sardinians and 2 Han Chinese; (ii) Ust’-Ishim, Altai Neanderthal, 2 Sardinians and 2 Han Chinese.

Relationship of Sunghir and Ust’-Ishim with Eurasians The aims of the model-based demographic analyses based on the SFS were to: (i)

understand the relationship between the ancient modern human samples (Sunghir and Ust’-Ishim) and Eurasian populations, and date their divergence; and (ii) test whether the ancient modern human samples (Sunghir/Ust’-Ishim) show any evidence of specific pulses of Neanderthal admixture in addition to what is found in Eurasians. We analysed the two ancient modern human individuals independently, but in the general description of the models we refer to Sunghir3 and Ust’-Ishim as “ancient” samples.

To investigate the relationship between ancient humans (Sunghir3 or Ust’-Ishim) and Eurasians, we considered two alternative models, representing two alternative population tree topologies: (i) split of ancient humans from the European branch (i.e. topology (Ancient,Europe),East Asia, Fig. S27A; or (ii) divergence of ancient humans from East Asians

(i.e. topology (Ancient,East Asia),Europe, Fig. S27B). Note that, under both models, the ancient sample can also split from the ancestral Eurasian branch and thus support a topology (Ancient, (Europe, East Asia)) (Fig. S27). In addition, we allowed for symmetric and constant migrations between Europeans and East Asians.

To investigate the number of major Neanderthal admixture events in the history of the ancient modern human samples (Sunghir/Ust’-Ishim), we considered two potential admixture events: (i) one admixture event specific to the ancient sample; (ii) one admixture event in the ancestral Eurasian population (Fig. S27). The search range for the admixture proportions was set between 10-6 and 10-1, such that values very close to zero (e.g., below 10-4) would suggest no admixture. Thus, the estimated admixture proportions can directly reveal if there is evidence for some extra pulse of admixture with Neanderthals in the ancient samtuples. We considered that the ancient modern humans (Sunghir/Ust’-Ishim) and the Eurasians admixed with two different ghost populations related to the Neanderthal Altai, one that admixed with the ancestral Eurasian population (Neanderthal-related Eurasians, N.R.E.), and another that admixed with the ancient human samples (Neanderthal-related ancient, N.R.A.) (Fig. S27). This modelling of admixture accounts for a likely genetic structure of Neanderthal populations, reflecting the fact that populations contributing to modern humans did not necessarily live in the Altai region, and therefore that admixture events might have occurred at different places. All divergence and admixture times were estimated assuming a constant mutation rate of 1.25e-8/gen/site (138) and a generation time of 29 years (19). Maximum likelihood parameters

We estimated the set of parameters that maximize the likelihood for each model by specifying the following search ranges: (i) all population effective sizes (Ne) could vary between 50 and 25,000 diploids, with an unbounded upper bound; (ii) the bottleneck duration was fixed to 100 generations and its size could vary between 5 and 2500 diploids (sizes larger than ~1250 correspond to mild or no bottleneck effect as the probability of coalescent events during that period becomes very small); (iii) the divergence time between Europeans and East Asians could have only occurred between 500 and 2,000 generations ago; (iv) the bottleneck could have be 10 to 4,000 generations older than the Eurasian split; (v) the migration rates could vary between 10-7 to 10-3 per generation per lineage (with a log-uniform search range such that small values could be better explored) based on previous estimates (96, 155). The divergence of the ancient sample could occur any time along the branch of Europeans/East Asians or along the ancestral Eurasian population branch, between the sampling time of the ancient human sample up to 5,000 generations ago. The sampling time of Sunghir was assumed to be 34 kya, which corresponds to 1,172 generations ago assuming 29 years/generation, whereas the sampling time of Ust’-Ishim was assumed to be 45 kya (15), corresponding to 1,530 generations ago. The Neanderthal admixture proportions could vary between 0.0001% and 10% (with a log-uniform search range) for the two admixture events. The search ranges of parameters related to the Neanderthal demography were based on the values reported in (7). The diploid effective sizes of Neanderthal could vary between 50 and 5,000 (again with open upper bounds). The sampling time for the Altai Neanderthal was set to 2,276 gen ago (~66 kya), which is the mid-point of the “absolute date calibration number 1” in (7). The divergence of the two ghost populations (N.R.E. and N.R.A.) from the Altai Neanderthal population could range from 2,280 (4 gen older than the sampling time of Altai Neanderthal) to an upper bound of 5,000 generations ago. The time of the admixture specific to the ancient humans could happen any time between the sampling time of

Sunghir 3 (34kya) or Ust’-Ishim (45 kya) and the split of ancient sample from Eurasians. The ancestral Eurasian admixture could have happened any time between the Eurasians split and the split time of the unsampled Neanderthal ghost populations. We also took into account the reported inbreeding coefficient of 0.125 for the Altai Neanderthal individual (7), and used it as the probability that two homologous chromosome are identical by descent (and thus coalesce immediately) at any locus.

The expected SFS used in the likelihood equations (1) and (2) for any given set of parameters was estimated from 500,000 coalescent simulations. For the original dataset of 294 blocks of 1Mb, we performed 100 optimizations runs starting from different initial conditions and selected the run leading to the highest likelihood to get parameter estimates.

Non parametric bootstrap analysis

We estimated confidence intervals for the model having the maximum likelihood by estimating parameters from 100 bootstrap datasets. All the settings for parameter estimation were the same as those used for the analyses of the original dataset, except that we only performed 10 optimization runs per bootstrap dataset, due to computational constraints. The 95% confidence intervals for each parameter were computed based on the percentile method (interval [Q0.025,Q0.975], where Qa is the a percentile of the bootstrap distribution) (156), as implemented in the R boot package. Results

We describe below the demographic history inferred from the two ancient modern human samples (Sunghir 3 and Ust’-Ishim) in separate sections.

Sunghir

The model where Sunghir 3 splits from the European lineage has a much higher (composite) likelihood (379 log10 units) than a model where it splits from East Asians (Table S24, Fig. S28). Even though the composite likelihoods cannot be strictly compared via Akaike information criterion (AIC) without computationally prohibitive simulations, the distribution of the likelihood values do not overlap (Fig. S28), clearly suggesting that the model with a divergence from Europe is much better supported by the data.

The parameter estimates obtained under the best model support a population tree topology ((SIII, Europe), East Asia) for all bootstrap datasets (Fig. S29), such that we can confidently say that the Sunghir 3 did not split from the ancestral Eurasian population branch, but directly from proto-Europeans ~38 kya (95% CI 35-43 kya). It suggests that the split of Sunghir from the lineage leading to present day Europeans occurred ~110 generations before the sampling time of Sunghir (34 kya), i.e. ~3.500 years (95% CI of up to 8.500 years).

Our results suggest a major Neanderthal admixture event of 2.51% (95% CI, 1.82-3.32) in the ancestors of Eurasians, shared by Sunghir and all present day Europeans and East Asians. We estimate this major event to have happened 55.5 kya (95% CI 51.8-63.2), thus very close to the time of divergence of Europeans and East Asians (~54.7 kya, 95%CI 49.8-58.0). Note that this major Neanderthal admixture is ~720 generations before the sampling time of Sunghir, which is comparable to the estimate obtained using the admixture tract length distribution (Supplementary Information 10). However, we also find evidence for an independent specific Neanderthal contribution of 0.37% (95% CI 0.0004-0.61%) into the Sunghir population, occurring ~50 generations before sampling, i.e. ~36 kya (95% CI 34-42), which could be part of the signal

detected with the LD-based methods and leading to a more recent admixture event. Furthermore, we find evidence for a very small effective size for Sunghir, with point estimates of Ne=297 diploids (95% CI, 158-901), which is consistent with the results of the IBD analyses (Figure 2; Supplementary Materials S7). All estimates and 95% CI for all inferred parameters are shown in Table S25. Ust’-Ishim

The SFS data analysis suggests that Ust’-Ishim diverged directly from the proto-East Asian population, as this model has a much higher likelihood (Table S26) than a model implying a divergence from the proto-Europeans. The likelihood of the best model is indeed 68 log10 units of likelihood better than the model with a divergence from Europeans, and there is no overlap in the likelihood distributions (Fig. S30), suggesting that a divergence from East Asians is indeed better supported by the data. The parameter estimates obtained under this model point to a population tree topology ((Ust’-Ishim, East Asia), Europe) (Fig. S30) for 99 out of the 100 bootstraps, again showing that most likely Ust’-Ishim did not diverge from the ancestors of Eurasians. We infer that Ust’-Ishim diverged from East Asians ~48 kya (95% CI 45-55 kya), so ~100 generations (~3000 years, with 95% CI of up to 10.000 years) before its sampling time (45 kya).

In close agreement with Sunghir results, we find a major Neanderthal admixture event in the ancestral of Eurasians of approximately ~2.5% (point estimate 2.38%, 95% CI, 1.86-2.92%). This major Neanderthal admixture event is estimated to have happened ~53 kya (95% CI 50.0-60.5 kya), close to the ~55 kya inferred with Sunghir data and with a broadly overlapping CI. The age of this event corresponds to approximately 290 generations before the sampling time of Ust’-Ishim, in close agreement with previously reported estimates of 232-430 generations based on the mean Neanderthal tract lengths (15). The Neanderthal admixture timing is also found to have happened just before the divergence of Europeans and East Asians (54.7 kya, 95%CI 49.8-58 kya), a time very similar to that estimated with Sunghir data. However, for Ust’-Ishim we find a stronger signal of specific Neanderthal contribution of 0.58% (95% CI 0.002-1.54) than for Sunghir, with an upper limit of the 95% CI reaching 1.5%. This extra Neanderthal admixture would have occurred ~47 kya (95% CI 44 – 51), i.e. ~70 generations before sampling. All estimates and 95% CI for all inferred parameters are shown in Table S27. Assessing the fit of the SFS

To visualize how well our model could reproduce the observed data, we compared the marginal distribution of the observed and expected SFS (Fig. S32, S33). Overall, we have a very good fit of the expected to the observed marginal SFS, suggesting that our model and the corresponding parameter estimates capture relevant aspects of the data. We also looked in more detail at the joint SFS (4 dimensions, 4D) to find the entries that could not be well explained by our model. Overall, even the entries with the worst fit are relatively well predicted/fitted (Fig. S34, S35).

S12. Neanderthal admixture S12.1. Neanderthal tract lengths

We used the sliding window approach of (11) to identify archaic admixture tracts and compare their length distributions in modern and ancient genomes. For a particular test individual, the method calls genomic tracts of likely archaic (Neanderthal / Denisova) origin by identifying runs of consecutive SNPs where the individual carries derived alleles selected to be informative about the respective archaic ancestry, using the following criteria:

- At least one copy of the derived allele observed in the test archaic - Only ancestral alleles observed in the other archaic - Only ancestral alleles observed in all African populations from the 1000 Genomes

Project Phase 3 dataset. - To ensure results are comparable between low and high coverage sets, we called tracts on

haploid genomes generated by randomly sampling a single allele at each site for the high coverage set. Archaic tracts were then defined as runs of derived alleles fulfilling the criteria above, allowing a maximum gap size of up to 200 kb between consecutive SNPs. To account for missing derived alleles due to differences in coverage between individuals and the random sampling at heterozygote genotypes, we allowed up to 13 ancestral alleles between consecutive derived alleles within a tract (corresponding to a binomial sampling probability of < 10-4). We furthermore required a tract to be at least 10 kb in length and contain a minimum of 3 SNPs. The fraction of Neanderthal ancestry per individual was obtained by the ratio of the sum of the total length of admixture tracts over the total length of the genome passing all filters.

Results for each individual are shown in Fig. S36. We find that average Neanderthal tract

length increases with age of the sample, as expected for individuals closer to the time of admixture. The longest tracts are observed in Ust’-Ishim, followed by Kostenki 14 and Sunghir individuals. We also observe a slight increase in total Neanderthal ancestry in those groups compared to modern European populations, which would be consistent with an additional admixture pulse or incomplete and ongoing purging of deleterious Neanderthal segments as suggested by the demographic modelling (Supplementary Materials S11).

S12.2. Dating of the admixture events using inferred admixture tracts.

To date the age of the Neanderthal admixture event, we inferred admixture tracts using the Hidden Markov Model (HMM) method of Seguin-Orlando et al(11). The strategy we use here is slightly different than in the original version. Instead of estimating the age directly by training the HMM on the data, we instead used fixed parameters for the HMM based on simulations and analytical predictions for a fixed introgression time and proportion. We then infer tracts using this model, and use parametric simulations to obtain an estimate of the introgression time from the inferred tracts. This procedure corrects for any biases associated with the tracts inferences because we use the exact same procedure for inferring tracts on the simulated and the real data. We prefer this method because it allows us to directly incorporate SNP filtering and biases due to inadequate modelling of background LD, in the inference procedure, thereby producing a more robust method for inferring introgression times.

We used an HMM with two states 1) sites heterozygous for the Neanderthal allele (SN) and 2) sites homozygous for the human alleles (SH). This construction allows us to analyse unphased data. We analyse that sites from the 1000 Genomes phase data that are variable in the CEU and YRI samples. We divide them into two categories: 1) sites in which the focal haplotype has a derived allele, Neanderthals have at least one derived allele, and YRI have zero derived alleles; 2) variable sites that do not meet the conditions in 1). Invariable sites in CEU and YRI, and sites for which we could not determine the conditions in 1), were considered missing data. The emission probabilities, estimated from simulated data, were 0.0162 and 0.000410 for observing a site of type 1) given the Neanderthal state and human state, respectively. The transition rates were computed from qHN = m(t - 1) per Morgan and qNH = (1-m)(t-1) per Morgan, where m=0.03 and t=728 generations. (Note: 728 generations = 1900 generations - sample age of 34kya in generations, where we use 29yrs/generations.) The prior of being in the Neanderthal state was set at m = 0.03. For both the real and simulated data, CEU and YRI sample sizes 85 and 88, respectively, were used to identify the variable sites within the union of the two populations. We then LD-pruned these sites at r2 = 0.7 via a sliding window approach (window size = 200kb). This procedure was applied to all 4 Sunghir samples.

We simulated data using SLiM (157) under demographic histories of Neanderthal

admixture into Sunghirs with varying admixture times. For each admixture time, 400 admixed chromosomes, with 100Mb/chromosome were simulated. A recombination rate of 1.3e-8, a mutation rate of 1.5e-8, and an admixture proportion of 0.03 were used. The inference procedure is robust to the assumed admixture proportion as long as it is relatively small. We fit an exponential distribution with parameter λ, truncated at 10 kb, to both the real and the simulated data. We then identified the simulated value of the admixture time that generates the same value of λ as observed in the real data. To obtain approximate confidence intervals (CIs) we block-bootstrap inferred introgression fragments.

The results are presented in Fig. S37. The mean value of λ for the real data was 1.05e-05

with an approximate 95% CI of [1.026 e-05, 1.067 e-05], resulting in an estimate of 770 generations before the age of the sample with an approximate 95% CI of [755, 786].

Fig. S1. (A) Map of Europe with UP sites 1) Sunghir, 2) Kostenki-Borshchevo sites, 3) Pavlov, 4) Dolní Věstonice, 5) Grub-Kranawetberg, 6) Předmostí, 7) Willendorf II, 8) Krems-Wachtberg, 9) Balzi Rossi Caves, 10) Arene Candide, 11) Molodova V, 12) Vorovonitsa I. Map based on GTOPO30 data (Data available from the U.S. Geological Survey). (B) Sunghir: map of the site showing location of trenches and burials (redrawn after(26)). (C) Sunghir: schematic section of squares 122-127 in Excavation Area III. 1) top soil, 2a and 2b) yellowish-tan loam and calcareous loess, 3) Cultural Layer, 4) yellowish sandy loam, 5) depositional hiatuses / erosional surface, 6) ice wedge cast filled with sediment (redrawn after (26)). (D) Lithic artefacts: Bifacially retouched points (1-5 redrawn after (26)). .

Fig. S2. The two main burials from Sunghir: (A) SI single burial, (B) SII-SIII juvenile double burial, including the SIV adult femur; (C) Personal ornaments (ivory beads and perforated carnivore teeth) from Sunghir. 1-3) Bead Type 1a, 4-7) Bead Type 1b, 9-10) Bead Type 2, 11-12) Bead Type 3, 14-16) Pendants made from perforated carnivore teeth (1-16 redrawn after(26)). Images: (A), open source photograph by José-Manuel Benito Álvarez, https://commons.wikimedia.org/wiki/File:Sunghir-tumba_paleol%C3%ADtica.jpg; (B), (C) after (26)

Fig. S3. Fragmentation and nucleotide misincorporation patterns along DNA reads for two DNA sequencing libraries constructed for the SIII_387 sample. (A) Patterns observed for a library built on a non-USER treated extract: SIII_387_NOT_USER_39_CGACCT. (B) Patterns observed for a library built on a USER-treated extract: SIII_387_USER_47_CATAGA.

Fig. S4. Bar plots indicating error rates in each substitution class for the sequenced individuals.

0.0000

0.0005

0.0010

0.0015

0.0020

A>C A>G A>T C>A C>G C>T G>A G>C G>T T>A T>C T>GNucleotide change

Erro

r rat

e

IndividualSISIISIIISIV

PopulationSunghir_UP

Fig. S5. Calibrated radiocarbon age of OxA-31755 (Sunghir 6).

Fig. S6. Calibrated radiocarbon age of OxA-X-2666-52 (Sunghir 5).

Fig. S7. Maximum likelihood tree of mitochondrial DNA sequences for Sunghir and other ancient Eurasian individuals.

Fig. S8. Maximum likelihood tree of Y chromosome sequences of Sunghir 3 and contemporary individuals from haplogroups A, B, C, D and G (127). The branch carrying mutation C-V20 found in one individual from Nepal and observed at low frequency in contemporary Southern Europeans is highlighted.

Fig. S9. Demographic model for coalescent simulations

Tdiv = 4,000 years

RefNE = 5,000

TestNE = x

Fig. S10. HBD tracts ≥ 2 Mb in archaic humans. Shown are tracts inferred for the Altai Neanderthal (top) and Denisovan (bottom). To note are long HBD tracts on chromosomes 14 and 21 in the Altai Neanderthal. In total, we infer 20 tracts with length ≥ 10 cM, in contrast to only 1 tract with ~ 10 cM in Denisova, consistent with the numbers reported in (7)

123456789

10111213141516171819202122

0 50 100 150 200 250Position [Mb]

Chr

omos

ome

AltaiNea / AltaiNea

123456789

10111213141516171819202122

0 50 100 150 200 250Position [Mb]

Chr

omos

ome

DenisovaPinky / Denisova

Fig. S11. Genotypes and HBD tracts on chromosome 18. Plot depicts genotypes (coded as 0/1/2 copies of the alternative allele) for individuals along chromosome 18. HBD tracts ≥ 2 Mb inferred by IBDseq are indicated by grey boxes with black lines. Symbol colour indicates transition (blue) or transversion (grey) variants

Fig. S12. HBD tracts inference comparison. Plot depicts HBD tracts for SIII, Ust’Ishim as well as archaic humans, inferred using IBDseq (blue) or a sliding window approach (red).

123456789

10111213141516171819202122

0 50 100 150 200 250Position [Mb]

Chr

omos

ome

groupibdseq

window

SIII

123456789

10111213141516171819202122

0 50 100 150 200 250Position [Mb]

Chr

omos

ome

groupibdseq

window

UstIshim

123456789

10111213141516171819202122

0 50 100 150 200 250Position [Mb]

Chr

omos

ome

groupibdseq

window

AltaiNea

123456789

10111213141516171819202122

0 50 100 150 200 250Position [Mb]

Chr

omos

ome

groupibdseq

window

Denisova

Fig. S13. Genotypes and HBD tracts on chromosome 4. Plot depicts genotypes (coded as 0/1/2 copies of the alternative allele) for respective individuals along chromosome 4. HBD tracts ≥ 2 Mb inferred by IBDseq and sliding window approaches are indicated by blue and red boxes, respectively. Symbol colour indicates transition (blue) or transversion (grey) variants

Fig. S14. HBD tract length distribution. Plot depicts the length distribution of HBD tracts ≥ 2Mb inferred using the sliding window approach in the respective individuals. An excess of very long tracts (≥ 10Mb) consistent with recent inbreeding can be clearly observed in the Altai Neanderthal.

0

10

20

30

40Al

taiN

ea

Den

isov

a

SIII

Ust

Ishi

m

Population

HBD

trac

t len

gth

(Mb)

Fig. S15. Pairwise IBD and local SFS between Sunghir 2 / Sunghir 3. Plot depicts the counts of IBS0 SFS entries for 100kb windows across the genome (blue dots). IBD tracts ≥ 4 Mb inferred by IBDseq are indicated by grey boxes.

Fig. S16. PCA with non-African reference populations.

Fig. S17. PCA with with West Eurasian reference populations.

Fig. S18. Shared genetic drift among ancient humans. Heatmap showing pairwise outgroup-f3 statistics of UP HGs and selected other ancient humans. Only individuals with > 50,000 SNPs covered are included in the plot.

Fig. S19. Outgroup-f3 statistics. Plot depicts the distributions of outgroup-f3 statistics for the study individuals and UP individuals from Kostenki. The five individuals with highest drift sharing in each panel are labelled.

Kostenki12

SI

SII

SIIISIV

Kostenki12Pavlov1

SII

SIII

SIV

Kostenki12 Rigney1

SI

SII

SIV

Pavlov1

SI

SIISIII

SIV

Kostenki12

Rigney1

SI

SIII

SIV

Kostenki12KremsWA3

SISII

SIII

SIII SIV

SI SII

Kostenki Kostenki12

Oase_UP

Ust_UP

Kostenki_

UPKostenki1

2_UP

Sunghir_UP

Goyet_U

PRo

mania_U

PPaglicci_UP

Krem

s_UP

Vestonice

_UP

Pavlov_UP

Ostuni_UP

Malta_U

PAfontovaGora_LP

ElMiron_LP

Goyet_LP

FranceHG

_LP

Villabruna_LP

GermanyHG_LP

Bichon_LP

FranceHG

_MLoschbour_M

Brana_M

HungaryHG_M

Motala_M

NordicH

G_M

NordicH

G_EN

EasternH

G_M

CaucasusHG

_LP

CaucasusHG

_MNa

tufian_LP

Levant_EN

Zagros_EN

Iran_EN

Boncuklu_EN

Tepecic_EN

Mentese_EN

Kumtepe_EN

Barcin_EN

Greece_EN

Hungary_EN

LBK_EN

Cardial_EN

Iberia_EN

Nordic_

EN

Oase_UP

Ust_UP

Kostenki_

UPKostenki1

2_UP

Sunghir_UP

Goyet_U

PRo

mania_U

PPaglicci_UP

Krem

s_UP

Vestonice

_UP

Pavlov_UP

Ostuni_UP

Malta_U

PAfontovaGora_LP

ElMiron_LP

Goyet_LP

FranceHG

_LP

Villabruna_LP

GermanyHG_LP

Bichon_LP

FranceHG

_MLoschbour_M

Brana_M

HungaryHG_M

Motala_M

NordicH

G_M

NordicH

G_EN

EasternH

G_M

CaucasusHG

_LP

CaucasusHG

_MNa

tufian_LP

Levant_EN

Zagros_EN

Iran_EN

Boncuklu_EN

Tepecic_EN

Mentese_EN

Kumtepe_EN

Barcin_EN

Greece_EN

Hungary_EN

LBK_EN

Cardial_EN

Iberia_EN

Nordic_

EN

0.10

0.15

0.20

0.25

0.30

0.35

0.10

0.15

0.20

0.25

0.30

0.35

0.10

0.15

0.20

0.25

0.30

0.35

Group

f 3

Fig. S20. Plot depicts quantile-quantile plots of expected and observed Z-scores for configuration

f4(Mbuti,Test)(SIII,X), for population X indicated at the respective panel. Test populations with

significant Z-scores (dashed lines) are indicated with larger plot symbols and their respective

label

Test closer to X

Test closer to SIII

Kostenki12

SI

SII

SIV

Vestonice16

Test closer to X

Test closer to SIII

Test closer to X

Test closer to SIII

Test closer to X

Test closer to SIII

SII

SII SIV

Kostenki SI

−2 −1 0 1 2 −2 −1 0 1 2

−10

−5

0

−10

−5

0

Expected

Obs

erve

d

Fig. S21. Base admixture graph model fit

b.paleoEur.qpgraph.graph :: Ust Vil MA1 Kos 0.009750 -0.002827 -0.012577 0.003852 -3.265

Mbuti_WGS

UstIshim

MA1

Kostenki

GoyetQ116-1 Villabruna

Vestonice16 ElMiron

Loschbour

R

non_African

139

10

X

21

455

Eurasia_W

10

X1

8

X2

4

435

Y1

203

X3

49

X4

83

pVes

84%

394

X5

90 354

X6

0

Y2

92

Y3

0

X7

100

Y4

19

pElm

54% pLos

9%

Y5

7

Y6

67 91%

16% 46%

287 316

23

Fig. S22. Admixture graph with best fit for SIII. Graph represents the best fit for SIII on the base model scaffold, equivalent to Figure 3C in the main text. We find three outlying f-statistics (max |Z| = 3.3) all involving the Villabruna sample. The model is based on 137,603 SNPs covered in all individuals

n02.b_s3.paleoEur.qpgraph.graph :: Ust Vil MA1 Kos 0.010998 -0.001939 -0.012936 0.003895 -3.321

Mbuti_WGS

UstIshim

MA1

Kostenki

GoyetQ116-1VillabrunaSIII

Vestonice16 ElMiron

Loschbour

R

non_African

141

10

X

22

459

Eurasia_W

11

X1

10

X2

4

435

X8

16

X3

50

X4

84

36

Y1

198 396

X5

87357

X6

0

pVes

84% Y2

94

Y3

0

X7

99

Y4

19

pElm

54% pLos

9%

Y5

7

Y6

67 91%

16% 46%

279 318

23

Fig. S23. Admixture graphs without fit of unadmixed SIII. Both topologies fail to fit with |Z| = 4.8 as worst f-statistic

n01.b_si.paleoEur.qpgraph.graph :: Ust Ves Kos SII 0.000000 0.018178 0.018178 0.003759 4.836

Mbuti_WGS

UstIshim

MA1

Kostenki GoyetQ116-1 VillabrunaSIII

Vestonice16 ElMiron

Loschbour

R

non_African

141

10

X

22

459

Eurasia_W

11

X1

14

X2

4

X8

0

Y1

203

X3

50

X4

82

43148

pVes

85%

395

X5

97 358

X6

0

Y2

90

Y3

0

X7

101

Y4

18

pElm

53% pLos

7%

Y5

6

Y6

68 93%

15% 47%

282 315

21

n03.b_s3.paleoEur.qpgraph.graph :: Ust Ves Kos SII 0.000000 0.018178 0.018178 0.003759 4.836

Mbuti_WGS

UstIshim

MA1

SIII

Kostenki GoyetQ116-1Villabruna

Vestonice16 ElMiron

Loschbour

R

non_African

141

10

X

22

459

Eurasia_W

11

X1

14

X2

3

48

X8

0

X3

50

X4

84

431

Y1

203 396

X5

84356

X6

0

pVes

85% Y2

94

Y3

0

X7

105

Y4

19

pElm

54% pLos

9%

Y5

6

Y6

66 91%

15% 46%

281 317

23

Fig. S24. Admixture graph without fit of admixed SIII. Topology fails with |Z| = 5.3 as worst f-statistic

x01.b_si.paleoEur.qpgraph.graph :: Kos ElM Ves SII 0.000000 -0.020390 -0.020390 0.003849 -5.297

Mbuti_WGS

UstIshim

MA1

Kostenki

GoyetQ116-1 Villabruna

Vestonice16 SIII

ElMiron

Loschbour

R

non_African

141

10

X

22

459

Eurasia_W

10

X1

10

X2

4

436

Y1

6

X3

50

X4

85

pVes

91%

396

X5

83 356

X6

0

Y2

96

Y3

0

X7

86

Y4

19

pElm

55% pLos

9%

Y5

0

Y6

66 91%

9% 45%

X8

7

421 40

321

23

Fig. S25. Admixture graphs including Kostenki 12. Plots showing admixture graphs attempting to fit Kostenki 12 onto the graph from Figure 1. (A), (B) graphs that fail to fit (|Z| > 3); (C), (D), Possible fits for Kostenki 12. All graphs indicate a closer relationship of Kostenki 12 to Sunghir than to either Kostenki 14 or Vestonice 16. Note that the lack of resolution due to the low number of SNPs results in no inferred contribution of the GoyetQ116-1 lineage to Loschbour.

Fig. S26. Population-specific drift estimates of Sunghir and selected ancient Eurasians. Panels show the estimate of genetic drift specific to the respective panel individual after its divergence with a set of modern and ancient individuals (indicated with coloured text).

Kost

enki

SI SIV

MA1SI

I SIII

MA1

SII

SI SIVSI

II

Ust

Ishi

mKo

sten

ki

SIISI SI

V

Ust

Ishi

m

MA1

MA1

SII SIV

Kost

enki

SI SIII

Ust

Ishi

mKo

sten

ki

MA1

SIISI SIVSI

II

Ust

Ishi

m

Bich

onSA

TPBi

chon SA

TPBi

chon

SATP

SATP

Bich

onSA

TP

Kare

lia

KK1

Losc

hbou

rLo

schb

our

Kare

liaKK1

Kare

liaKK1

Losc

hbou

rKa

relia

KK1

Losc

hbou

rLo

schb

our

Kare

lia

KK1

Bon0

02W

C1

Bar8

Klei

10St

uttg

art

Stut

tgar

t

WC

1

Bar8

Bon0

02

Klei

10

Bon0

02 Stut

tgar

t

WC

1

Bar8

Klei

10

Bon0

02 Stut

tgar

t

WC

1

Bar8

Klei

10

Bon0

02 Stut

tgar

tWC

1

Klei

10

SS60

0446

7SS

6004

468

SS60

0447

4

SS60

0447

9SS

6004

469

SS60

0447

2

SS60

0447

6SS

6004

477

SS60

0447

8

SS60

0446

7SS

6004

468

SS60

0447

2

SS60

0447

4

SS60

0447

9SS

6004

469

SS60

0447

6SS

6004

477

SS60

0447

8

SS60

0446

7SS

6004

468

SS60

0447

2

SS60

0447

4

SS60

0447

9SS

6004

476

SS60

0446

9

SS60

0447

7SS

6004

478

SS60

0446

7

SS60

0447

4

SS60

0447

9

SS60

0446

8 SS60

0446

9

SS60

0447

2

SS60

0447

6SS

6004

477

SS60

0447

8

SS60

0446

7SS

6004

468

SS60

0447

2

SS60

0447

4 SS60

0447

9SS

6004

469

SS60

0447

6SS

6004

477

SS60

0447

8

Upp

erPa

leol

ithic

Late

Pale

olith

ic

Mes

olith

ic

Neo

lithi

c

Mod

ern

UstIshim

KostenkiSIII

BichonBar8

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

0.0

0.1

0.2

0.3

Individual

Drif

t uni

ts

PopulationSardinian2

French2

Dai2

Han2

Mixe2

Karitiana2

Australian

Papuan2

Ust_UP

Kostenki_UP

Sunghir_UP

Malta_UP

Bichon_LP

CaucasusHG_M

CaucasusHG_LP

WesternHG_M

EasternHG_M

Zagros_EN

Boncuklu_EN

Barcin_EN

Greece_LN

LBK_EN

Fig. S27. Schematic representation of the models tested in the present study to infer the relationship of the ancient modern human samples (SIII: Sunghir3; UI: Ust’-Ishim) with Eurasians. a) model where ancient humans split from proto-Europeans. b) model where ancient humans split from proto-East Asians. Under both models, ancient humans can also split from the ancestors of Eurasians (as indicated by dashed lines), depending on the inferred divergence times of the split of the ancient sample (TD_Anc) and the divergence of Eurasians (TD_EuAs). The split of the ancient humans can also happen before or after the admixture with Neanderthals. Neanderthal admixture was modelled as having occurred from two unsampled populations related to Altai Neanderthals, N.R.E. (Neanderthal-related contribution to the ancestral Eurasian population) and N.R.A. (Neanderthal-related contribution to ancient human population), indicated by light blue arrows. The models also include a bottleneck in the ancestral Eurasian population, which could represent the Out of Africa event, and symmetric migrations between Europeans and East Asians.

Fig. S28. Comparison of the Log likelihood for the models with the topology ((SIII, Eur), EAs), “Europe”, with the model with topology (SIII, (Eur, EAs)), “East Asia”. A) Log-likelihood values computed including the number of monomorphic sites in the likelihood function (Lfull in equation 1); B) Log-likelihood values computed only from the polymorphic sites (LSFS in eq. 2). Distributions were obtained from 100 expected SFS approximated with 106 coalescent simulations. These distributions reflect the noise resulting from approximating the likelihood with coalescent simulations.

Fig. S29. Schematic representation of the parameter estimates obtained for the Sunghir3 individual under the best model with topology ((SIII, Europe), East Asia). Point estimates are shown in bold, and 95% confidence intervals are shown within square brackets. Note that due to the re-scaling of the time parameters (see Statistical inference section), the sampling times can be slightly younger than 34 kya. Times of divergence in years are obtained assuming a generation time of 29 years and a mutation rate of 1.25e-8/gen/site.

Fig. S30. Comparison of the Log likelihood for the models with the topology (UI, (Eur, EAs)), “Europe”, with the model with topology ((UI, EAs), Eur), “East Asia”. A) Log-likelihood values computed including the number of monomorphic sites in the likelihood function (Lfull in equation 1); B) Log-likelihood values computed only based on the polymorphic sites (LSFS in equation 2). Distributions were obtained from 100 expected SFS approximated with 106 coalescent simulations. These distributions reflect the noise resulting from approximating the likelihood with coalescent simulations.

Fig. S31. Schematic representation of the parameter estimates obtained for the Ust’-Ishim individual under the best model with topology ((UI, East Asia), Europe). Point estimates are shown in bold, and 95% confidence intervals are shown within square brackets. Note that due to the re-scaling of the time parameters (see Statistical inference section), the sampling times can be slightly younger than 45kya. Times of divergence in years are obtained assuming a generation time of 29 years and a mutation rate of 1.25e-8/gen/site.

Fig. S32. Comparison of the marginal observed and marginal expected SFS for Sunghir3. The x-axis shows the derived allele frequencies (allele counts) and the y-axis shows the number of SNPs with a given frequency (in log10 scale). The expected SFS was obtained as the average of 100 simulated SFSs (each approximated with 106 coalescent simulations), according to the maximum-likelihood parameter estimates obtained with the original dataset for Sunghir.

Fig. S33. Comparison of the marginal observed and marginal expected SFS for Ust’-Ishim. The x-axis shows the derived allele frequencies (allele counts) and the y-axis shows the number of SNPs with a given frequency (in log10 scale). The expected SFS was obtained as the average of 100 simulated SFSs (each approximated with 106 coalescent simulations), according to the maximum-likelihood parameter estimates obtained with the original dataset for Ust’-Ishim.

Fig. S34. Comparison of the multidimensional joint observed and expected SFS for the 26 entries showing the worst fit (out of the 225 entries) in the multidimensional joint SFS with Sunghir3. We defined an arbitrary threshold to define entries with the worst fit, by selecting those that exhibit a difference between the expected and observed SFS larger than 100 log10Lhood units (i.e. |(miLog10(pi))- (miLog10(mi/L)|>100, where mi is the observed counts at the i-th entry, pi is the expected SFS at the i-th entry and L is the total number of polymorphic sites). Each column corresponds to one entry of the SFS, coded as n,g,s,h (from bottom to top) as the frequency of the derived allele in Altai Neanderthal (n), Sunghir3 (g), Sardinians (s), and Han Chinese (h). A) Comparison of the expected and observed counts (y-axis) for each entry. B) Comparison of the relative fit, defined as the relative number of SNP counts for a given entry (Relative fit= #expSNPs/ #obsSNPs, y-axis). Expected SFS were obtained as the average of 100 simulated SFSs (approximated with 106 coalescent simulations), according to the parameter estimates obtained under the best model for the Sunghir3 original dataset. Error bars correspond to the 0.01 and 0.99 quantiles of the 100 simulated SFSs.

Fig. S35. Comparison of the multidimensional joint observed and expected SFS for the 26 entries showing the worst fit (out of the 225 entries) in the multidimensional joint SFS with Ust’-Ishim. We defined an arbitrary threshold to define entries with the worst fit, by selecting those that exhibit a difference between the expected and observed SFS larger than 100 log10Lhood units (i.e. |(miLog10(pi))- (miLog10(mi/L)|>100, where mi is the observed counts at the i-th entry, pi is the expected SFS at the i-th entry and L is the total number of polymorphic sites). Each column corresponds to one entry of the SFS, coded as n,u,s,h (from bottom to top) as the frequency of the derived allele in Altai Neanderthal (n), Ust’-Ishim (u), Sardinians (s), and Han Chinese (h). A) Comparison of the expected and observed counts (y-axis) for each entry. B) Comparison of the relative fit, defined as the relative number of SNP counts for a given entry (Relative fit= #expSNPs/ #obsSNPs, y-axis). Expected SFS were obtained as the average of 100 simulated SFSs (approximated with 106 coalescent simulations), according to the parameter estimates obtained under the best model for the Ust’-Ishim original dataset. Error bars correspond to the 0.01 and 0.99 quantiles of the 100 simulated SFSs.

Fig. S36. Genomic segments of putative archaic origins. Shown are the distributions of the average length and overall genomic fraction of genomic segments putatively originating from (A), Altai Neanderthal and (B), Denisovan in modern and ancient human genomes.

Fig. S37. Neanderthal admixture time estimate for SIII using tract length distribution.

Table S1. Radiocarbon dates on materials from Sunghir. Radiocarbon ages calibrated using IntCal13 (32) and OxCal 4.2 (33). LabNo 14C years

BP

CalBP68.2% CalBP95.4%

Material

C:Nra

tio

Metho

d

Pretreatmen

t

Commen

ts

Reference

Cultu

ralLayer

GIN-14 14,600±600 18,517-17,028 19,206-16,200 Bone a 1

Lye-1058 19,780±80 23,943-23,685 24,063-23,558 Carbona-

ceoussoil

a excavationII,Horizon5 1

GIN-9585 20,360±900 25,603-23,575 26,610-22,551 Bone a Mammothfemur,mixed 2

GIN-326a 21,800 ±

1000

27,267-25,186 28,222-24,013 Charcoal a Hearth 1

GIN-326b 22,500±600 27,330-26,205 27,836-25,720 Charcoal a 1

GIN-8998 23,600±600 28,394-27,305 29,026-26,606 Bone a Mammothbone,mixed 2

Gro-5446 24,430±400 28,840-28,020 29,346-27,751 collagen a Rangifer 1

Gro-5425 25,500±200 29,905-29,306 30,283-29,081 Charcoal a 3.2mdepth 1

GIN-9001 25,770±600 30,650-29,384 31,028-28,768 Bone a 7horsefrags.,mixed 2

GIN-8995 26,300±260 30,870-30,329 31,016-29,881 Bone a Mammothfemur,H3 2

GIN-9034 26,300±300 30,896-30,278 31,043-29,796 Bone a 5horsefrags.,H3-4 2

GIN-9030 26,600±300 31,056-30,616 31,234-30,233 Bone a Mammothfemur,H1 2

GIN-9031 26,630±280 31,648-31,170 32,220-31,021 Bone a Mammothfemur,H2 2

GIN-9035 26,900±260 31,160-30,847 31,320-30,670 Bone a Rangifervertebra 2

GIN-9591 27,000±320 31,243-30,866 31,473-30,644 Bone a Mammothdiaphysisfrag 2

GIN-9027 27,200±400 31,433-30,908 32,041-30,618 Bone a Mammothulna 2

GIN-9586 27,200±500 31,578-30,805 32,543-30,460 Bone a Mammothfemur,H3 2

GIN-9036 27,260±500 31,639-30,835 32,610-30,550 Bone a Rangifervertebra 2

GIN-9033 27,400±400 31,605-30,992 32,371-30,790 Bone a 6horsefrags. 2

OxA-9039 27,460±310 31,524-31,083 32,032-30,887 Bone 3.5 b X Mammothdiaphysis,H4 3

GIN-5880 27,700±500 32,175-31,110 32,969-30,904 Bone a Mammothhumerus 2

GIN-9588 27,800±600 32,442-31,161 33,330-30,910 Bone a Mammothvertebra 2

GIN-8997 28,000±250 32,131-31,416 32,623-31,280 Bone a Mammothfemur 2

GIN-9029 28,000±300 32,221-31,401 32,744-31,241 Bone a Mammothfemur,H2 2

GIN-8999 28,120±170 32,183-31,564 32,572-31,437 Bone a Mammothhumerus,H3 2

GIN-8996 28,130±370 32,490-31,496 33,057-31,251 Bone a Mammothfemur 2

GIN-9032 28,350±200 32,615-31,870 32,902-31,590 Bone a Mammothfemur,H1 2

GIN-9028 28,800±240 33,367-32,638 33,617-32,115 Bone a Mammothulna,H3 2

OxA-15755 29,450±189 33,841-33,511 34,006-33,275 bone 3.2 b Y Mammoth 4

OxA-15752 29,640±189 33,959-33,655 34,125-33,498 Bone 3.1 b Y Mammoth 4

OxA-X-2395-8 30,100±400 34,510-33,851 34,878-33,557 Bone 5.1 b Z Mammoth 4

SI

AA-36473 19,160±270 23,436-22,764 23,731-22,476 Bone b vertfrag. 5

OxA-9036 22,930±200 27,466-27,087 27,623-26,752 Bone b Y 3

KIA-27006 27,050±210 31,212-30,953 31,340-30,810 Bone 3.1 b Y 6

OxA-X-2464-12 28,890±430 33,616-32,506 33,875-31,770 Bone 5.0 b Z 7SII

OxA-9037 23,830±220 28,107-27,699 28,411-37,568 Bone 3.5 b X 3

OxA-15753 25,020±120 29,222-28,871 29,413-28,740 Bone 3.3 b Y 4

AA-36475 26,200±640 30,955-29,744 31,362-28,996 Bone b Leftribs 5

AA-36474 27,210±710 32,087-30,640 33,148-29,919 Bone b Rightribs 5

OxA-X-2395-6 30,100±550 34,655-33,740 35,283-33,185 Bone 5.0 b Z 4

SIII

OxA-9038 24,100±240 28,387-27,889 28,645-27,735 Bone 3.4 b X 3

OxA-15754 24,830±110 28,982-28,708 29,175-28,575 Bone 3.2 b Y 4

OxA-15751 25,430±160 29,740-29,280 30,076-29,053 Bone 3.2 b Y 4

KIA-27007 26,000±410 30,569-30,049 30,711-29,772 Bone 3.5 b Y 6

AA-36476 26,290±640 30,947-29,735 31,353-38,990 Bone b 5

OxA-X-2395-7 30,000±550 34,596-33,665 35,154-33,031 Bone 5.0 b Z 5

SIV OxA-X-2462-52 29,820±280 34,165-33,706 34,485-33,499 Bone 5.1 b Z 7

SV OxA-X-2666-52 26,042±182 30,644-30,074 30,780-29,746 Bone 3.4 b Z 8

OxA-31755 884±23 894-744 905-732 Bone 3.3 b Y 8

SVI

OxA-X2653-36 925±29 906-850 and

831-795

924-728 Bone 5.0 b Z 8

Unit1

GIN-15 16,200±400 20,015-19,065 20,553-18,716 Soil a Organicsoil 1

GIN-16 20,540±120 24,943-24,513 25,150-24,360 Soil a Organicsoil,1mdepth 1

Table S2. Frequency of main lithic artefact categories from Sunghir (after (26)).

Category n

Big fragments (raw material fragments) 9417

Cores and core-like fragments 604

Flakes 38902

Blades 929

Crested blades 98

Burin spalls 104

Tools 1869

Total 51923

Table S3. Types of ivory beads from Sunghir. Based on (26) and (36). Typeof

ivory

bead

Attributes Size Productionsequence Locationwithinthesite

Type 1

(a,b)

Semi-rectangular,

with sharp or

round edges,

thinned at the

waist. May or not

beperforated.

~ 13 x 9 x 5

mm,

Perforation

Ø:~6mm

Dominantbeadproductionsequence

atthesite.Asmallivoryrodisincised

all around at regular intervals and

snapped into pieces forming the

bead blanks. The blanks are then

thinned at the waist on both sides

(Type 1b), and predominantly

perforated centrally through the

thinnedarea(Type1a).

AlltypesofbeadsatSunghir

were found associated with

the 3 buried skeletons (SI,

SII,SIII),aswellaswithinthe

CulturalLayer.

Type2 Small, rounded

bead with central

perforation

~ 6 x 4 x 2.5

mm

Generally similar to Type 1a, but

smallerandrounded.

The centrally waisted beads

(Type 1) found with SI are ~

1/3 larger than those found

withthejuvenilesSII-SIII.

Type3 Sub-rectangular or

oval beads, pierced

ononeside

12x8x4mm The blanks are flat or moderately

rounded and thinner one the side of

theperforation.

Type3beadsarerare.

Table S4. Upper Palaeolithic multiple burials Site

#ofin

dividu

als Age Position of the

bodiesOchre Associatedartefacts

Sunghir 2 Sub-adult Extended onback, head-to-head

Yes, concentratedon the top of thebody

Multiple, including >10,000 ivory beads, ivorybracelets,pendants,spears,figurineanddiscs,aswellasArcticfoxcanines

DolníVěstoniceII

3 Adults Extended, DV-13 and DV-15on their back,DV-14 prone;head to theSouth

Yes, concentratedaroundtheskullsofall3 individualsandthe pelvic area ofthe middleindividualDV-15

*burnedwood inandaroundthegrave,aswellas large quantities of ash and charcoal in theburialpitandsurroundingarea;* All three individuals were found with piercedcarnivoreteethandinsomecasesbeadsontheircrania(DV-13:tworowsofperforatedfoxcanines(n=20) with one flat drop-shaped ivory pendantoneachsideontheforehead,heldtogetherbyacrust of ochre; DV-14: drop-shaped/oval-shapedivorypendantsonforehead(ornexttohead)andthree wolf canines; DV-15: four pierced foxcanines) (65,71). Melanopsis sp. shells werefoundinlowquantitiesscatteredwithinthetripleburial,noneofwhichwasperforated,butmostofwhichwerepartlyburned(71).

Grotta deiFanciulli

2 Adolecentmale andadultfemale

Excavators argue that themale was buried firstand the graveopened later again to add female(44).

BarmaGrande

3 An adultand 2adolescents

Barma Grande2 on its back,Barma Grande3and4ontheirleft in slightlyflexedposition

All three bodiesstained with redochre

BovidfemurplacedunderheadofMarmaGrande3(thecentralindividual),largescraperundertheheadofBarmaGrande4 (left individual); BarmaGrande 3 partly covered by the other two, soprobably first in burial. Personal ornamentsinclude perforated shells of Cyclope nerita andincised deer canines, fish vertebrae and bonependants on heads and chests, as well as longflintblades(41,44).

Krems-Wachtberg

2 Peri-natal Both bodiescovered in redochre

Thegravewascoveredbyamodifiedmammothscapula and contained a tusk fragment (notmodified). Individual A had a string of > 30 flatoval shaped ivory beads with a proximalperforationpositionedaroundthepelvis(73).

Table S5. Summary of individuals newly reported in this study.

Contamination Age

Genome mtDNA (%) mtDNA Y (cal BP (95%))Sunghir 1 1.11 690 M 0.90 ± 0.21 U8c C1a2 33,875-31,770

Sunghir 2 4.08 2,506 M 0.33 ± 0.07 U2 C1a2 35,283-33,185

Sunghir 3 10.75 7,467 M 0.54 ± 0.07 U2 C1a2 35,154-33,031

Sunghir 4 3.87 674 M 0.54 ± 0.08 U2 C1a2 34,485-33,499

Sunghir 5 0.001 0.07 - - - - 30,780-29,746*

Sunghir 6 4.19 2,408 M 13.1 ± 0.31 W3a1 I2a1b2 730-850*

* New radiocarbon dates from this study

CoverageIndividual Sex

Haplogroup

Table S6. DNA extractions and sequencing libraries. Sample IDs of different biological samples of the respective individuals are in bold typeface.

digestiontime libraryID USERtreament PCRcycles #ReadPairs %Uniq.Endo. Cov(X) Size(bp)SI_389 leftfemurextraction1 130mg low-speeddrilling o/n SI_389_USER_39_CGACCT USER 9 14,082,390 0.26% 0.0006 49.9

SI_389_USER_40_TGACGT USER 9 16,259,604 0.25% 0.0006 50.1SI_389_USER_41_TGTCTG USER 9 8,162,686 0.26% 0.0003 49.7SI_389_USER_42_ACGTGC USER 9 13,502,813 0.26% 0.0005 50.1SI_389_NOT_USER_43_TGATCC notUSER 10 12,235,742 0.52% 0.0010 52.1SI_389_NOT_USER_44_CTCTAG notUSER 10 9,998,461 0.55% 0.0008 52.1

SI_388 molarrootextraction1 46mg microdismembrator o/n SI_388_NOT_USER_13_CTATCA notUSER 7+7 24,334,398 11.12% 0.0426 55.4

SI_388_NOT_USER_15_TGATCG notUSER 8+7 13,870,044 0.52% 0.0013 56.9SI_388_USER_14_CGTATA USER 7+7 1,072,142,789 1.42% 0.1923 50.5SI_388_USER_5_ACAGTG USER 13 276,179,272 3.57% 0.1235 50.6SI_388_USER_17_CGCTAT_B USER 13 317,997,627 8.54% 0.3315 49.3SI_388_USER_21_ACATAC USER 11 489,081,387 7.64% 0.4106 48.0

extraction2 pellet 24h SI_388_USER_17_CGCTAT USER 8+7 21,407,515 0.40% 0.0013 51.7SI_388_USER_36_TGCAGG USER 12 12,876,285 0.49% 0.0009 52.8SI_388_USER_40_TGACGT USER 14 40,554,662 0.34% 0.0021 53.4

extraction3 160mg microdismembrator o/n SI_388_USER_29_TGATGC USER 6 66,962,934 0.16% 0.0012 49.0extraction4 pellet 24h SI_388_USER_50_TTGAAC USER 8 53,559,629 0.07% 0.0004 49.2extraction5 pellet 96h SI_388_USER_34_GAGATA USER 14 61,880,094 0.25% 0.0015 46.2

Total 2,525,088,332 Coverage 1.11

SII_383 pre-molarrootextraction1 220mg microdismembrator o/n SII_383_NOT_USER_21_ACATAC notUSER 8 25,065,069 0.51% 0.0017 50.3

SII_383_USER_46_CACGAA USER 8 40,581,208 0.64% 0.0027 46.3extraction2 pellet 24h SII_383_NOT_USER_29_TGATGC notUSER 9 26,469,031 0.49% 0.0019 53.1

SII_383_USER_33_GACCGG USER 10 20,400,026 0.63% 0.0014 48.0extraction3 pellet 72h SII_383_NOT_USER_37_ACTGCC notUSER 9 29,433,099 0.31% 0.0016 62.9

SII_383_USER_41_TGTCTG USER 9 25,152,908 0.43% 0.0013 53.5extraction4 120mg microdismembrator o/n SII_383_USER_31_GACACT USER 10 871,559,087 8.92% 1.2031 60.9

SII_383_USER_7_CAGATC USER 10 205,826,925 10.58% 0.3496 61.4SII_383_USER_37_ACTGCC USER 10 491,102,957 10.16% 0.7878 61.3SII_383_USER_41_TGTCTG_B USER 9 453,321,993 10.08% 0.7238 61.3SII_383_USER_34_GAGATA USER 10 245,034,226 11.60% 0.4556 60.5SII_383_USER_38_GCAACG USER 8 263,319,852 11.90% 0.5195 62.0

extraction5 pellet 24h SII_383_USER_32_ACGCAT USER 12 79,304,899 0.87% 0.0099 59.1extraction6 pellet 96h SII_383_USER_35_CTGACA USER 14 74,867,213 0.54% 0.0081 63.4

SII_384 righttibiaextraction1 245mg microdismembrator o/n SII_384_NOT_USER_22_TGAGCC notUSER 7 24,886,162 0.15% 0.0005 46.3

SII_384_USER_26_TGTGAG USER 8 27,522,190 0.13% 0.0004 44.0extraction2 pellet 24h SII_384_NOT_USER_48_ACAGTC notUSER 10 27,114,036 0.43% 0.0017 47.8

SII_384_USER_34_GAGATA USER 10 26,355,256 0.32% 0.0011 45.4extraction3 pellet 72h SII_384_NOT_USER_38_GCAACG notUSER 10 26,552,711 0.78% 0.0037 50.6

SII_384_USER_42_ACGTGC USER 10 22,323,097 0.41% 0.0015 48.1

Total 3,006,191,945 Coverage 4.08

SIII_386 molarrootextraction1 186mg microdismembrator o/n SIII_386_NOT_USER_2_CGATGT notUSER 10 35,354,408 15.17% 0.1108 70.9

SIII_386_USER_4_TGACCA USER 12 475,924,285 11.28% 0.8559 64.2SIII_386_USER_6_GCCAAT USER 10 129,342,366 13.89% 0.3032 65.9SIII_386_USER_18_TGAACA USER 11 399,082,136 15.52% 1.0256 64.6SIII_386_USER_22_TGAGCC USER 9 1,227,785,202 20.82% 3.5785 62.6SIII_386_USER_39_CGACCT USER 9 43,259,541 16.86% 0.1259 64.9SIII_386_USER_35_CTGACA USER 10 53,767,122 16.36% 0.1511 64.9

extraction2 pellet 24h SIII_386_NOT_USER_6_GCCAAT notUSER 11 27,868,423 23.87% 0.1502 72.3SIII_386_USER_8_ACTTGA_A USER 11 361,077,714 17.36% 1.0411 64.1SIII_386_USER_7_CAGATC USER 11 71,255,369 12.47% 0.1667 67.9SIII_386_USER_19_GTATCT USER 11 311,942,249 21.32% 1.2167 66.3

extraction3 pellet 96h SIII_386_NOT_USER_9_GATCAG notUSER 11 15,639,374 17.84% 0.0598 70.1SIII_386_USER_11_GGCTAC USER 12 219,524,250 14.78% 0.5648 64.8SIII_386_USER_8_ACTTGA_B USER 11 150,598,355 14.92% 0.3927 65.2SIII_386_USER_20_CAGCTA USER 11 289,572,606 16.89% 0.8706 65.0

SIII_385 molarrootextraction1 127mg microdismembrator o/n SIII_385_NOT_USER_1_ATCACG notUSER 9 27,946,416 4.30% 0.0246 70.1

SIII_385_USER_3_TTAGGC USER 9 31,637,433 5.03% 0.0268 63.3extraction2 pellet 24h SIII_385_NOT_USER_5_ACAGTG notUSER 9 20,552,135 3.21% 0.0134 68.6

SIII_385_USER_7_CAGATC USER 9 22,942,288 3.63% 0.0137 63.2extraction3 pellet 96h SIII_385_NOT_USER_18_TGAACA notUSER 7+6 25,214,246 0.44% 0.0020 62.2

SIII_385_USER_20_CAGCTA USER 7+6 16,879,491 0.79% 0.0020 57.9

SIII_387 rightfemur

extraction1 240mg microdismembrator o/n SIII_387_NOT_USER_23_AGCATG notUSER 6 17,270,341 0.68% 0.0018 58.2SIII_387_USER_47_CATAGA USER 6 17,059,384 0.62% 0.0015 55.7

extraction2 pellet 24h SIII_387_NOT_USER_31_GACACT notUSER 9 29,245,560 2.37% 0.0129 61.7SIII_387_USER_35_CTGACA USER 9 39,699,926 2.12% 0.0147 58.8

extraction3 pellet 72h SIII_387_NOT_USER_39_CGACCT notUSER 9 24,505,261 3.05% 0.0152 59.0SIII_387_USER_43_TGATCC USER 8 23,826,435 2.50% 0.0099 54.3

Total 4,108,772,316 Coverage 10.75

SIV_392 femurextraction1 300mg microdismembrator o/n SIV_392_NOT_USER_37_ACTGCC notUSER 8 14,574,063 4.02% 0.0088 61.0

SIV_392_NOT_USER_38_GCAACG notUSER 8 9,090,494 4.03% 0.0055 61.1SIV_392_USER_33_GACCGG USER 9 308,784,436 4.25% 0.1813 58.7SIV_392_USER_34_GAGATA USER 9 312,322,004 4.21% 0.1813 58.9SIV_392_USER_35_CTGACA USER 9 325,248,952 4.24% 0.1855 58.2SIV_392_USER_36_TGCAGG USER 9 347,339,549 4.11% 0.1970 58.7

extraction2 pellet 48h SIV_392_USER_45_GAGAAG USER 9 437,753,766 4.52% 0.2595 56.6SIV_392_USER_46_CACGAA USER 9 267,017,368 5.33% 0.1755 56.2SIV_392_USER_47_CATAGA USER 9 293,943,863 4.71% 0.1954 58.2SIV_392_USER_48_ACAGTC USER 9 415,687,778 4.67% 0.2761 58.5SIV_392_USER_49_CATCGT USER 10 374,041,258 4.63% 0.2216 56.2SIV_392_USER_50_TTGAAC USER 10 292,100,250 4.72% 0.1805 57.0

extraction3 pellet 72h SIV_392_NOT_USER_10_TAGCTT notUSER 8 25,739,846 7.72% 0.0382 64.9SIV_392_USER_12_CTTGTA USER 8 958,446,953 6.29% 0.9555 58.7SIV_392_USER_9_GATCAG USER 7 608,065,194 5.59% 0.5661 60.6SIV_392_USER_23_AGCATG USER 8 213,780,389 6.39% 0.2389 60.1

Total 5,203,936,163 Coverage 3.87

SVI_393 mandibulaextraction1 215mg microdismembrator o/n SVI_393_NOT_USER_24_CGATGA notUSER 9 24,782,443 6.70% 0.0286 58.8

SVI_393_USER_28_TCTCGC USER 10 687,995,114 4.31% 0.4469 55.8SVI_393_USER_10_TAGCTT USER 10 330,670,431 5.02% 0.2525 56.1SVI_393_USER_24_CGATGA USER 10 137,679,076 5.58% 0.1199 56.0

extraction2 pellet 24h SVI_393_NOT_USER_32_ACGCAT notUSER 13 17,876,132 18.18% 0.0571 60.5SVI_393_USER_36_TGCAGG USER 13 1,189,784,309 2.25% 0.4056 56.2SVI_393_USER_11_GGCTAC USER 13 326,029,231 5.77% 0.2923 56.4SVI_393_USER_26_TGTGAC USER 14 71,287,306 12.59% 0.1450 56.6

extraction3 pellet 72h SVI_393_NOT_USER_40_TGACGT notUSER 12 34,668,295 12.29% 0.0743 58.5SVI_393_USER_44_CTCTAG USER 12 734,797,928 3.38% 0.3921 55.8SVI_393_USER_12_CTTGTA USER 11 312,321,710 5.22% 0.2578 55.8SVI_393_USER_28_TCTCGC_B USER 12 220,895,691 7.41% 0.2695 56.2

extraction4 215mg microdismembrator o/n SVI_393_USER_49_CATCGT USER 8 207,805,443 9.43% 0.3597 62.0extraction5 pellet 24h SVI_393_USER_33_GACCGG USER 12 101,587,927 21.53% 0.4223 63.3extraction6 pellet 96h SVI_393_USER_36_TGCAGG_B USER 12 166,185,095 20.59% 0.6645 63.3

Total 4,564,366,131 Coverage 4.19

digestiontime libraryID USERtreament PCRcycles #ReadPairs %Uniq.Endo. Cov(X) Size(bp)SI_389 leftfemurextraction1 130mg low-speeddrilling o/n SI_389_USER_39_CGACCT USER 9 14,082,390 0.26% 0.0006 49.9

SI_389_USER_40_TGACGT USER 9 16,259,604 0.25% 0.0006 50.1SI_389_USER_41_TGTCTG USER 9 8,162,686 0.26% 0.0003 49.7SI_389_USER_42_ACGTGC USER 9 13,502,813 0.26% 0.0005 50.1SI_389_NOT_USER_43_TGATCC notUSER 10 12,235,742 0.52% 0.0010 52.1SI_389_NOT_USER_44_CTCTAG notUSER 10 9,998,461 0.55% 0.0008 52.1

SI_388 molarrootextraction1 46mg microdismembrator o/n SI_388_NOT_USER_13_CTATCA notUSER 7+7 24,334,398 11.12% 0.0426 55.4

SI_388_NOT_USER_15_TGATCG notUSER 8+7 13,870,044 0.52% 0.0013 56.9SI_388_USER_14_CGTATA USER 7+7 1,072,142,789 1.42% 0.1923 50.5SI_388_USER_5_ACAGTG USER 13 276,179,272 3.57% 0.1235 50.6SI_388_USER_17_CGCTAT_B USER 13 317,997,627 8.54% 0.3315 49.3SI_388_USER_21_ACATAC USER 11 489,081,387 7.64% 0.4106 48.0

extraction2 pellet 24h SI_388_USER_17_CGCTAT USER 8+7 21,407,515 0.40% 0.0013 51.7SI_388_USER_36_TGCAGG USER 12 12,876,285 0.49% 0.0009 52.8SI_388_USER_40_TGACGT USER 14 40,554,662 0.34% 0.0021 53.4

extraction3 160mg microdismembrator o/n SI_388_USER_29_TGATGC USER 6 66,962,934 0.16% 0.0012 49.0extraction4 pellet 24h SI_388_USER_50_TTGAAC USER 8 53,559,629 0.07% 0.0004 49.2extraction5 pellet 96h SI_388_USER_34_GAGATA USER 14 61,880,094 0.25% 0.0015 46.2

Total 2,525,088,332 Coverage 1.11

SII_383 pre-molarrootextraction1 220mg microdismembrator o/n SII_383_NOT_USER_21_ACATAC notUSER 8 25,065,069 0.51% 0.0017 50.3

SII_383_USER_46_CACGAA USER 8 40,581,208 0.64% 0.0027 46.3extraction2 pellet 24h SII_383_NOT_USER_29_TGATGC notUSER 9 26,469,031 0.49% 0.0019 53.1

SII_383_USER_33_GACCGG USER 10 20,400,026 0.63% 0.0014 48.0extraction3 pellet 72h SII_383_NOT_USER_37_ACTGCC notUSER 9 29,433,099 0.31% 0.0016 62.9

SII_383_USER_41_TGTCTG USER 9 25,152,908 0.43% 0.0013 53.5extraction4 120mg microdismembrator o/n SII_383_USER_31_GACACT USER 10 871,559,087 8.92% 1.2031 60.9

SII_383_USER_7_CAGATC USER 10 205,826,925 10.58% 0.3496 61.4SII_383_USER_37_ACTGCC USER 10 491,102,957 10.16% 0.7878 61.3SII_383_USER_41_TGTCTG_B USER 9 453,321,993 10.08% 0.7238 61.3SII_383_USER_34_GAGATA USER 10 245,034,226 11.60% 0.4556 60.5SII_383_USER_38_GCAACG USER 8 263,319,852 11.90% 0.5195 62.0

extraction5 pellet 24h SII_383_USER_32_ACGCAT USER 12 79,304,899 0.87% 0.0099 59.1extraction6 pellet 96h SII_383_USER_35_CTGACA USER 14 74,867,213 0.54% 0.0081 63.4

SII_384 righttibiaextraction1 245mg microdismembrator o/n SII_384_NOT_USER_22_TGAGCC notUSER 7 24,886,162 0.15% 0.0005 46.3

SII_384_USER_26_TGTGAG USER 8 27,522,190 0.13% 0.0004 44.0extraction2 pellet 24h SII_384_NOT_USER_48_ACAGTC notUSER 10 27,114,036 0.43% 0.0017 47.8

SII_384_USER_34_GAGATA USER 10 26,355,256 0.32% 0.0011 45.4extraction3 pellet 72h SII_384_NOT_USER_38_GCAACG notUSER 10 26,552,711 0.78% 0.0037 50.6

SII_384_USER_42_ACGTGC USER 10 22,323,097 0.41% 0.0015 48.1

Total 3,006,191,945 Coverage 4.08

SIII_386 molarrootextraction1 186mg microdismembrator o/n SIII_386_NOT_USER_2_CGATGT notUSER 10 35,354,408 15.17% 0.1108 70.9

SIII_386_USER_4_TGACCA USER 12 475,924,285 11.28% 0.8559 64.2SIII_386_USER_6_GCCAAT USER 10 129,342,366 13.89% 0.3032 65.9SIII_386_USER_18_TGAACA USER 11 399,082,136 15.52% 1.0256 64.6SIII_386_USER_22_TGAGCC USER 9 1,227,785,202 20.82% 3.5785 62.6SIII_386_USER_39_CGACCT USER 9 43,259,541 16.86% 0.1259 64.9SIII_386_USER_35_CTGACA USER 10 53,767,122 16.36% 0.1511 64.9

extraction2 pellet 24h SIII_386_NOT_USER_6_GCCAAT notUSER 11 27,868,423 23.87% 0.1502 72.3SIII_386_USER_8_ACTTGA_A USER 11 361,077,714 17.36% 1.0411 64.1SIII_386_USER_7_CAGATC USER 11 71,255,369 12.47% 0.1667 67.9SIII_386_USER_19_GTATCT USER 11 311,942,249 21.32% 1.2167 66.3

extraction3 pellet 96h SIII_386_NOT_USER_9_GATCAG notUSER 11 15,639,374 17.84% 0.0598 70.1SIII_386_USER_11_GGCTAC USER 12 219,524,250 14.78% 0.5648 64.8SIII_386_USER_8_ACTTGA_B USER 11 150,598,355 14.92% 0.3927 65.2SIII_386_USER_20_CAGCTA USER 11 289,572,606 16.89% 0.8706 65.0

SIII_385 molarrootextraction1 127mg microdismembrator o/n SIII_385_NOT_USER_1_ATCACG notUSER 9 27,946,416 4.30% 0.0246 70.1

SIII_385_USER_3_TTAGGC USER 9 31,637,433 5.03% 0.0268 63.3extraction2 pellet 24h SIII_385_NOT_USER_5_ACAGTG notUSER 9 20,552,135 3.21% 0.0134 68.6

SIII_385_USER_7_CAGATC USER 9 22,942,288 3.63% 0.0137 63.2extraction3 pellet 96h SIII_385_NOT_USER_18_TGAACA notUSER 7+6 25,214,246 0.44% 0.0020 62.2

SIII_385_USER_20_CAGCTA USER 7+6 16,879,491 0.79% 0.0020 57.9

SIII_387 rightfemur

extraction1 240mg microdismembrator o/n SIII_387_NOT_USER_23_AGCATG notUSER 6 17,270,341 0.68% 0.0018 58.2SIII_387_USER_47_CATAGA USER 6 17,059,384 0.62% 0.0015 55.7

extraction2 pellet 24h SIII_387_NOT_USER_31_GACACT notUSER 9 29,245,560 2.37% 0.0129 61.7SIII_387_USER_35_CTGACA USER 9 39,699,926 2.12% 0.0147 58.8

extraction3 pellet 72h SIII_387_NOT_USER_39_CGACCT notUSER 9 24,505,261 3.05% 0.0152 59.0SIII_387_USER_43_TGATCC USER 8 23,826,435 2.50% 0.0099 54.3

Total 4,108,772,316 Coverage 10.75

SIV_392 femurextraction1 300mg microdismembrator o/n SIV_392_NOT_USER_37_ACTGCC notUSER 8 14,574,063 4.02% 0.0088 61.0

SIV_392_NOT_USER_38_GCAACG notUSER 8 9,090,494 4.03% 0.0055 61.1SIV_392_USER_33_GACCGG USER 9 308,784,436 4.25% 0.1813 58.7SIV_392_USER_34_GAGATA USER 9 312,322,004 4.21% 0.1813 58.9SIV_392_USER_35_CTGACA USER 9 325,248,952 4.24% 0.1855 58.2SIV_392_USER_36_TGCAGG USER 9 347,339,549 4.11% 0.1970 58.7

extraction2 pellet 48h SIV_392_USER_45_GAGAAG USER 9 437,753,766 4.52% 0.2595 56.6SIV_392_USER_46_CACGAA USER 9 267,017,368 5.33% 0.1755 56.2SIV_392_USER_47_CATAGA USER 9 293,943,863 4.71% 0.1954 58.2SIV_392_USER_48_ACAGTC USER 9 415,687,778 4.67% 0.2761 58.5SIV_392_USER_49_CATCGT USER 10 374,041,258 4.63% 0.2216 56.2SIV_392_USER_50_TTGAAC USER 10 292,100,250 4.72% 0.1805 57.0

extraction3 pellet 72h SIV_392_NOT_USER_10_TAGCTT notUSER 8 25,739,846 7.72% 0.0382 64.9SIV_392_USER_12_CTTGTA USER 8 958,446,953 6.29% 0.9555 58.7SIV_392_USER_9_GATCAG USER 7 608,065,194 5.59% 0.5661 60.6SIV_392_USER_23_AGCATG USER 8 213,780,389 6.39% 0.2389 60.1

Total 5,203,936,163 Coverage 3.87

SVI_393 mandibulaextraction1 215mg microdismembrator o/n SVI_393_NOT_USER_24_CGATGA notUSER 9 24,782,443 6.70% 0.0286 58.8

SVI_393_USER_28_TCTCGC USER 10 687,995,114 4.31% 0.4469 55.8SVI_393_USER_10_TAGCTT USER 10 330,670,431 5.02% 0.2525 56.1SVI_393_USER_24_CGATGA USER 10 137,679,076 5.58% 0.1199 56.0

extraction2 pellet 24h SVI_393_NOT_USER_32_ACGCAT notUSER 13 17,876,132 18.18% 0.0571 60.5SVI_393_USER_36_TGCAGG USER 13 1,189,784,309 2.25% 0.4056 56.2SVI_393_USER_11_GGCTAC USER 13 326,029,231 5.77% 0.2923 56.4SVI_393_USER_26_TGTGAC USER 14 71,287,306 12.59% 0.1450 56.6

extraction3 pellet 72h SVI_393_NOT_USER_40_TGACGT notUSER 12 34,668,295 12.29% 0.0743 58.5SVI_393_USER_44_CTCTAG USER 12 734,797,928 3.38% 0.3921 55.8SVI_393_USER_12_CTTGTA USER 11 312,321,710 5.22% 0.2578 55.8SVI_393_USER_28_TCTCGC_B USER 12 220,895,691 7.41% 0.2695 56.2

extraction4 215mg microdismembrator o/n SVI_393_USER_49_CATCGT USER 8 207,805,443 9.43% 0.3597 62.0extraction5 pellet 24h SVI_393_USER_33_GACCGG USER 12 101,587,927 21.53% 0.4223 63.3extraction6 pellet 96h SVI_393_USER_36_TGCAGG_B USER 12 166,185,095 20.59% 0.6645 63.3

Total 4,564,366,131 Coverage 4.19

Table S7. Sex determination Individual N ChrY+ChrX N ChrY RY 95% CI Assignment Sunghir 1 1,933,152 169,340 0.088 0.0872-0.0880 XY Sunghir 2 5,696,964 515,102 0.090 0.0902-0.0907 XY Sunghir 3 14,601,304 1,313,571 0.090 0.0898-0.0901 XY Sunghir 4 5,673,831 502,690 0.089 0.0884-0.0888 XY Sunghir 6 6,121,859 537,116 0.088 0.0875-0.0880 XY

Table S8. HPLC gradient for the separation of underivatised amino acids on Primesep A column with MilliQ™ deionised water (eluent A) and 0.3% phosphoric acid diluted with MilliQ™ deionised water (eluent B).

Time (min) % eluent A % eluent B

0 - 20 100 0

20 - 21 Linear gradient Linear gradient

21 - 320 0 100

Table S9. Radiocarbon determinations and analytical data from Sunghir 6. CRA is conventional radiocarbon age, expressed in years BP (98). PCode refers to pretreatment code; AF* is ultrafiltered collagen (* with pre-solvent wash), HYP denotes the extraction of hydroxyproline from hydrolysed bone collagen (99, 158). Collagen yield represents the weight of gelatin or ultrafiltered gelatin in milligrams. %Yld is the percent yield of extracted collagen as a function of the starting weight of the bone analysed. %C is the carbon present in the combusted gelatin. Stable isotope ratios are expressed in ‰ relative to vPDB with a mass spectrometric precision of ±0.2‰ (159). CN is the atomic ratio of carbon to nitrogen and is acceptable if it ranges between 2.9—3.5 in the case of collagen or ~5.0 in the case of hydroxyproline.

OxA CRA Error P

code

Used

(mg)

Yield

(mg)

%Yld %C 13C

(‰)

15N

(‰)

CN

atomic

ratio

31755 884 23 AF* 750 68.9 9.2 44.6 -17.6 10.8 3.3

X-2653-36 925 29 Hyp na na na 41.8 -16.3 15.6 5.0

Table S10. Radiocarbon determinations and analytical data from Sunghir 5. CRA is conventional radiocarbon age, expressed in years BP (98). PCode MAA refers to pretreatment based on HPLC to collect specific amino acids. See caption for Table S9 for details of the other analytical parameters.

OxA CRA Error P

code

Bone

used

(mg)

Yield

(mg)

%Yld %C 13C

(‰)

15N

(‰)

CN

atomic

ratio

X-2666-52 25240 160 MAA 1330 16.44 1.2 34.9 -17.9 12.9 3.4

Table S11. High-coverage ancient individuals and their respective coverage on the nuclear genome

Individual Coverage Reference

SIII 10.8 This study Ust'-Ishim 42.0 (15) Loschbour 22.0 (120) Stuttgart 19.0 (120)

NE1 22.1 (81) BR2 21.2 (81)

Bichon 9.5 (108) KK1 15.4 (108)

Clovis 14.4 (144) Denisova 31.0 (13)

Altai Neandertal 52.0 (7)

Table S12. Sunghir 1 Y chromosome mutations

SNP Haplogroup Position Mutation Coverage Mutations covered

Fraction of derived mutations

M130 C 2,734,854 C->T 1

5 1.00 P255 C 8,685,038 G->A 1 P260 C 17,286,006 A->C 1

Page85 C 14,924,643 G->T 1 V183 C 14,263,271 G->A 2 F3393 C1 23,023,974 C->A 1 1 1.00

Z28847 C1a2 14,400,573 C->A 2 97 0.02

Z28864 C1a2 15,852,360 A->G 1 Z28888 C1a2a 16,870,467 G->A 1 30 0.03

Table S13. Sunghir 2 Y chromosome mutations

SNP Haplogroup Position Mutation Coverage Mutations covered

Fraction of derived

mutations M130 C 2,734,854 C->T 1

9 0.88

P184 C 7,218,128 T->C 2 P255 C 8,685,038 G->A 1 P260 C 17,286,006 A->C 3

Page85 C 14,924,643 G->T 2 V183 C 14,263,271 G->A 2 V232 C 7,629,098 T->C 4 V77 C 17,947,542 C->T 2

F3393 C1 23,023,974 C->A 3 1 1.00 CTS11043 C1a 22,914,979 G->T 4 1 1.00

Z28011 C1a2 21,143,412 G->A 1

186 0.03 Z28785 C1a2 7,682,488 A->T 1 Z28831 C1a2 9,848,996 T->C 3 Z28864 C1a2 15,852,360 A->G 1 Z28925 C1a2 18,246,830 C->T 1

Table S14. Sunghir 3 Y chromosome mutations

SNP Haplogroup Position Mutation Coverage Mutations covered

Fraction of derived

mutations M130 C 2,734,854 C->T 8

11 0.91

M216 C 15,437,564 C->T 4 P184 C 7,218,128 T->C 3 P255 C 8,685,038 G->A 4 P260 C 17,286,006 A->C 7

Page85 C 14,924,643 G->T 3 V183 C 14,263,271 G->A 8 V199 C 2,772,928 C->A 5 V232 C 7,629,098 T->C 8 V77 C 17,947,542 C->T 5

F3393 C1 23,023,974 C->A 6 1 1.00 CTS11043 C1a 22,914,979 G->T 8 1 1.00

Z28011 C1a2 21,143,412 G->A 2

210 0.03

Z28785 C1a2 7,682,488 A->T 3 Z28831 C1a2 9,848,996 T->C 3 Z28847 C1a2 14,400,573 C->A 8 Z28864 C1a2 15,852,360 A->G 5 Z28925 C1a2 18,246,830 C->T 4 Z28965 C1a2 21,265,778 G->A 7

Table S15. Sunghir 4 Y chromosome mutations

SNP Haplogroup Position Mutation Coverage Mutations covered

Fraction of derived

mutations M130 C 2,734,854 C->T 1

9 0.89

M216 C 15,437,564 C->T 5 P184 C 7,218,128 T->C 2 P255 C 8,685,038 G->A 2 P260 C 17,286,006 A->C 2

Page85 C 14,924,643 G->T 4 V199 C 2,772,928 C->A 5 V232 C 7,629,098 T->C 3 F3393 C1 23,023,974 C->A 5 1 1.00

CTS11043 C1a 22,914,979 G->T 2 1 1.00 Z28011 C1a2 21,143,412 G->A 2

185 0.02 Z28831 C1a2 9,848,996 T->C 3 Z28847 C1a2 14,400,573 C->A 3 Z28864 C1a2 15,852,360 A->G 2

Table S16. Expected values for different kinship estimators

Relationship k0 k1 k2 KING (range) R1 (range) Parent-offspring 0 1 0 0.25 [0.177, 0.354] 0.5

Full siblings 0.25 0.5 0.25 0.25 [0.177, 0.354] (0.5, 0.769] 2nd degree 0.5 0.5 0 0.125 [0.0884, 0.177] [0.167, 0.444] 3rd degree 0.75 0.25 0 0.0625 [0.0442, 0.0884] [0.071, 0.421] Unrelated 1 0 0 0 [0, 0.4]

Table S17. Inbreeding coefficient from ngsRelate using CEU frequencies All SNPs Transversion only Individual F N Sites F N Sites Sunghir 1 0.085 3,711,079 0.090 1,196,720 Sunghir 2 0.065 5,662,736 0.073 1,821,218 Sunghir 3 0.067 5,879,199 0.074 1,892,081 Sunghir 4 0.063 5,602,157 0.075 1,801,884

Table S18. Relatedness estimation from ngsRelate using CEU frequencies

k0 k1 k2 N SNPs k0 k1 k2 N SNPsSunghir 1 Sunghir 2 0.655 0.111 0.234 3,633,438 0.641 0.115 0.244 1,169,904Sunghir 1 Sunghir 3 0.546 0.336 0.118 3,698,981 0.540 0.335 0.125 1,192,149Sunghir 1 Sunghir 4 0.727 0.017 0.256 3,617,160 0.716 0.016 0.268 1,164,474Sunghir 2 Sunghir 3 0.467 0.416 0.117 5,617,929 0.463 0.411 0.126 1,806,435Sunghir 2 Sunghir 4 0.554 0.367 0.079 5,413,590 0.546 0.367 0.087 1,739,717Sunghir 3 Sunghir 4 0.524 0.386 0.089 5,562,379 0.523 0.379 0.099 1,788,406Sunghir 1 Kostenki 14 0.614 0.328 0.058 3,146,009 0.610 0.325 0.065 1,004,225Sunghir 2 Kostenki 14 0.679 0.244 0.077 4,664,592 0.676 0.237 0.087 1,485,257Sunghir 3 Kostenki 14 0.671 0.266 0.062 4,776,308 0.667 0.265 0.069 1,521,340Sunghir 4 Kostenki 14 0.649 0.290 0.061 4,629,391 0.643 0.289 0.068 1,474,093

Bichon Loschbour 0.880 0.095 0.025 4,740,413 0.558 0.345 0.097 1,853,463Villabruna Loschbour 0.601 0.334 0.065 1,039,144 0.596 0.312 0.092 321,447La Brana Loschbour 0.639 0.293 0.068 5,031,224 0.634 0.287 0.079 1,595,166

between LP/Meso sites

Individual pair All SNPs Transversion only Site relationship

between UP sites

within site

Table S19. Inbreeding coefficients from ngsRelate using Paleolithic / Mesolithic frequencies All SNPs Transversion only Individual F N Sites F N Sites Sunghir 1 0.009 576,690 0.010 112,536 Sunghir 2 0.001 792,778 0.001 155,007 Sunghir 3 0.000 799,470 0.001 156,420 Sunghir 4 0.002 791,231 0.004 154,724

Table S20. Relatedness estimation from ngsRelate using Paleolithic / Mesolithic frequencies

k0 k1 k2 N SNPs k0 k1 k2 N SNPsSunghir 1 Sunghir 2 0.875 0.099 0.025 572,149 0.875 0.106 0.027 111,586Sunghir 1 Sunghir 3 0.927 0.033 0.039 576,595 0.927 0.034 0.041 112,506Sunghir 1 Sunghir 4 0.898 0.085 0.017 571,128 0.898 0.092 0.022 111,387Sunghir 2 Sunghir 3 0.801 0.146 0.052 792,594 0.801 0.144 0.056 154,950Sunghir 2 Sunghir 4 0.975 0.001 0.024 784,567 0.975 0.001 0.028 153,316Sunghir 3 Sunghir 4 0.853 0.141 0.005 791,060 0.853 0.135 0.005 154,677Sunghir 1 Kostenki 14 1.000 0.000 0.000 510,845 1.000 0.000 0.000 99,089Sunghir 2 Kostenki 14 1.000 0.000 0.000 699,416 1.000 0.000 0.000 135,752Sunghir 3 Kostenki 14 1.000 0.000 0.000 704,852 1.000 0.000 0.000 136,863Sunghir 4 Kostenki 14 1.000 0.000 0.000 698,122 1.000 0.000 0.000 135,510

Bichon Loschbour 0.724 0.248 0.027 795,957 0.724 0.247 0.029 155,765Villabruna Loschbour 0.892 0.036 0.072 603,105 0.892 0.039 0.093 114,845La Brana Loschbour 0.899 0.058 0.043 734,757 0.899 0.064 0.045 142,228

Individual pair Site relationship

between UP sites

between LP/Meso sites

All SNPs Transversion only

within site

Table S21. Relatedness estimation from Relate (CEU frequencies)

k0 k1 k2 N SNPsSunghir 1 Sunghir 2 0.425 0.575 0.000 46,199Sunghir 1 Sunghir 3 0.444 0.556 0.000 46,596Sunghir 1 Sunghir 4 0.453 0.547 0.000 46,084Sunghir 2 Sunghir 3 0.386 0.581 0.033 51,295Sunghir 2 Sunghir 4 0.430 0.570 0.000 50,422Sunghir 3 Sunghir 4 0.853 0.141 0.005 791,060Sunghir 1 Kostenki 14 0.630 0.370 0.000 44,911Sunghir 2 Kostenki 14 0.504 0.496 0.000 48,892Sunghir 3 Kostenki 14 0.487 0.513 0.000 49,282Sunghir 4 Kostenki 14 0.483 0.517 0.000 48,669

Bichon Loschbour 0.554 0.446 0.000 50,221Villabruna Loschbour 0.616 0.384 0.000 50,586La Brana Loschbour 0.533 0.467 0.000 52,269

0

between LP/Meso sites

Individual pair All SNPs Site relationship

between UP sites

within site

Table S22. Relatedness estimation from Relate (Paleolithic / Mesolithic frequencies)

k0 k1 k2 N SNPsSunghir 1 Sunghir 2 1.000 0.000 0.000 6,506Sunghir 1 Sunghir 3 0.969 0.031 0.000 6,519Sunghir 1 Sunghir 4 1.000 0.000 0.000 6,500Sunghir 2 Sunghir 3 0.775 0.219 0.006 7,436Sunghir 2 Sunghir 4 1.000 0.000 0.000 7,417Sunghir 3 Sunghir 4 0.864 0.136 0.000 7,427Sunghir 1 Kostenki 14 1.000 0.000 0.000 6,164Sunghir 2 Kostenki 14 1.000 0.000 0.000 7,050Sunghir 3 Kostenki 14 1.000 0.000 0.000 7,064Sunghir 4 Kostenki 14 1.000 0.000 0.000 7,045

Bichon Loschbour 0.932 0.068 0.000 7,109Villabruna Loschbour 0.940 0.060 0.000 7,125La Brana Loschbour 0.807 0.193 0.000 7,313

between LP/Meso sites

Individual pair All SNPs Site relationship

between UP sites

within site

Table S23. IBS-based relatedness estimates at 1240K SNP sites

KING R1 ratio degree Individual 1 Individual 2 Individual 1 Individual 2 Individual 1 Individual 2I0736 I0854 Barcin_EN 0.236 0.370 1 2.25 1.57 N1a1a1a N1a1a1aI0114 I0117 Central_EBA 0.227 0.374 1 1.07 1.90 I3a I3a I2a2I0744 I1097 Barcin_EN 0.151 0.283 2 2.39 2.13 J1c11 W1-T119C G2a2b2a G2a2b2aI0012 I0014 Motala_M 0.077 0.248 3 1.76 2.71 U2e1 U5a2d I2c

Sunghir 1 Sunghir 2 Sunghir_UP -0.042 0.196 unrelated 1.11 4.08 U8c U2 C1a2 C1a2Sunghir 1 Sunghir 3 Sunghir_UP -0.053 0.196 unrelated 1.11 10.75 U8c U2 C1a2 C1a2Sunghir 1 Sunghir 4 Sunghir_UP -0.060 0.198 unrelated 1.11 3.87 U8c U2 C1a2 C1a2Sunghir 2 Sunghir 3 Sunghir_UP 0.034 0.237 unrelated 4.08 10.75 U2 U2 C1a2 C1a2Sunghir 2 Sunghir 4 Sunghir_UP -0.014 0.211 unrelated 4.08 3.87 U2 U2 C1a2 C1a2Sunghir 3 Sunghir 4 Sunghir_UP -0.001 0.216 unrelated 10.75 3.87 U2 U2 C1a2 C1a2

mtDNA haplogroup Y haplogroupGroupIndividual pair Relatedness Coverage

Table S24. Log-likelihood values obtained for the models with SIII. Note that under both models the topology (SIII, (Eur, EAs)) could be favoured, depending on the estimated times of divergence of Sunghir3 (TD_Anc) and times of divergence of Europeans and East Asians (TD_EuAs) (see Figure S27).

Model

Possible Topologies

Favored topology

No. of parameters

Estimated Log10(Lhood)

Divergence from

Europeans

(SIII, Eur), EAs if TD_Anc<TD_EuAs

SIII, (Eur, EAs) if TD_Anc>TD_EuAs

(SIII, Eur), EAs

22

-1,320,802

Divergence from

East Asians

(SIII, EAs), Eur if TD_Anc<TD_EuAs

SIII, (Eur, EAs) if TD_Anc>TD_EuAs

SIII, (Eur, EAs)

22

-1,321,181

Table S25. Point estimates and 95% confidence intervals for the parameters of the best model obtained for SIII with topology ((SIII, Europe), East Asia). Confidence intervals were calculated according to the percentile method. Point estimates correspond to the parameters inferred with the original data set. Times of divergence in years are obtained assuming a generation time of 29 years and a mutation rate of 1.25e-8/gen/site.

95% Confidence intervals

Parameter Point estimate Lower limit Upper limit

Effective sizes (number of diploids)

Ne ancestral archaics/humans 17270 16745 17832

Ne Neanderthal Altai 1003 60 1030

Ne Neanderthal contributing to Eurasia (N.R.E.) 10942 3538 22091

Ne Neanderthal contributing to Sunghir (N.R.A.) 12865 3056 22390

Ne Neanderthal ancestral population 5296 4218 7591

Ne Ancient (Sunghir3) 296 158 901

Ne Europeans (Sardinians) 9942 8365 11111

Ne East Asians (Han Chinese) 6243 5591 6976

Ne ancestral Eurasians 2194 741 21605

Ne bottleneck 338 241 2175

Ne ancestral modern humans 14905 12985 17554

Times of divergence (in kya)

Divergence modern humans/Neanderthal 545 514 562

Time for divergence Eurasians 55 50 58

Time for divergence Sunghir 38 35 43

Time for divergence N.R.E. from Altai

Neanderthal 129

68 135

Time for divergence N.R.A. from Altai

Neanderthal 125

68 132

Time for Bottleneck in ancestral Eurasian 60 59 138

Time Neanderthal admixture into Eurasians 55 52 63

Time Neanderthal admixture into Sunghir 36 34 42

Admixture proportions (in %)

Admixture N.R.A. Sunghir 0.3664 0.0004 0.6118

Admixture N.R.E. Eurasians 2.5056 1.8232 3.3208

Migration rates (forward in time)

Migration from East Asia into Europe 0.0155 0.0036 0.0700

Migration from Europe into East Asia 0.0098 0.0024 0.0440

Table S26. Log-likelihood values obtained for the models including Ust’-Ishim. Note that under both models the topology (UI, (Eur, EAs)) could be favoured, depending on the estimated times of divergence of Ust’-Ishim (TD_Anc) and times of divergence of Europeans and East Asians (TD_EuAs) (see Figure S27).

Model

Possible Topologies

Favored topology

No. of parameters

Estimated Log10(Lhood)

Divergence from

Europeans

(UI, Eur), EAs if TD_Anc<TD_EuAs

UI, (Eur, EAs) if TD_Anc>TD_EuAs

(UI, Eur), EAs

22

-1,330,727

Divergence from

East Asians

(UI, EAs), Eur if TD_Anc<TD_EuAs

UI, (Eur, EAs) if TD_Anc>TD_EuAs

UI, (Eur, EAs)

22

-1,330,659

Table S27. Point estimates and 95% confidence intervals for the parameters of the best model obtained for Ust’-Ishim with topology ((UI, Europe), East Asia). Confidence intervals were calculated according to the percentile method. Point estimates correspond to the parameters inferred with the original data set. Times of divergence in years are obtained assuming a generation time of 29 years and a mutation rate of 1.25e-8/gen/site.

95% Confidence intervals

Parameter Point estimate Lower limit Upper limit

Effective sizes (number of diploids)

Ne ancestral archaics/humans 17343 16842 17756

Ne Neanderthal Altai 810 70 1101

Ne Neanderthal contributing to Eurasia (N.R.E.) 17872 4428 22307

Ne Neanderthal contributing to Ust’-Ishim

(N.R.A.) 16466 3046 25853

Ne Neanderthal ancestral population 4830 4235 6637

Ne Ancient (Ust’-Ishim) 1203 253 7098

Ne Europeans (Sardinians) 8746 7741 10073

Ne East Asians (Han Chinese) 5903 5296 6562

Ne ancestral Eurasians 1423 389 22004

Ne bottleneck 1685 287 2240

Ne ancestral modern humans 14370 13195 17328

Times of divergence (in kya)

Divergence modern humans/Neanderthal 539 518 562

Time for divergence Eurasians 52 49 57

Time for divergence Ust’-Ishim 48 45 55

Time for divergence N.R.E. from Altai

Neanderthal 105

70 133

Time for divergence N.R.A. from Altai

Neanderthal 126

70 138

Time for Bottleneck in ancestral Eurasian 63 53 96

Time Neanderthal admixture into Eurasians 53 50 61

Time Neanderthal admixture into Ust’-Ishim 47 44 51

Admixture proportions (in %)

Admixture N.R.A. Ust’-Ishim 0.5826 0.0022 1.5393

Admixture N.R.E. Eurasians 2.3772 1.8578 2.9147

Migration rates (forward in time)

Migration from East Asia into Europe 0.0124 0.0034 0.5604

Migration from Europe into East Asia 0.0084 0.0023 0.3651

Table S28. Outgroup – f3 statistics f3(Mbuti;A,X). Available online separate excel files.

Table S29. f4 statistics f4(Mbuti,P2;P3,P4). Available online separate excel files.

References

1. J.-P. Bocquet-Appel, P.-Y. Demars, L. Noiret, D. Dobrowsky, Estimates of Upper Palaeolithic meta-population size in Europe from archaeological data. J. Archaeol. Sci. 32, 1656–1668 (2005). doi:10.1016/j.jas.2005.05.006

2. M. M. Lahr, R. A. Foley, in Neanderthals and Modern Humans in the European Landscape During the Last Glaciation: Archaeological Results of the Stage 3 Project (McDonald Institute for Archaeological Research, Cambridge, 2003), pp. 241–256.

3. E. Trinkaus, A. P. Buzhilova, M. B. Mednikova, M. V. Dobrovolskaya, The People of Sunghir: Burials, Bodies, and Behavior in the Earlier Upper Paleolithic (Oxford Univ. Press, 2014).

4. See supplementary materials.

5. A. Marom, J. S. O. McCullagh, T. F. G. Higham, A. A. Sinitsyn, R. E. M. Hedges, Single amino acid radiocarbon dating of Upper Paleolithic modern humans. Proc. Natl. Acad. Sci. U.S.A. 109, 6878–6881 (2012). doi:10.1073/pnas.1116328109 Medline

6. S. Nalawade-Chavan, J. McCullagh, R. Hedges, New hydroxyproline radiocarbon dates from Sungir, Russia, confirm early Mid Upper Palaeolithic burials in Eurasia. PLOS ONE 9, e76896 (2014). doi:10.1371/journal.pone.0076896 Medline

7. K. Prüfer, F. Racimo, N. Patterson, F. Jay, S. Sankararaman, S. Sawyer, A. Heinze, G. Renaud, P. H. Sudmant, C. de Filippo, H. Li, S. Mallick, M. Dannemann, Q. Fu, M. Kircher, M. Kuhlwilm, M. Lachmann, M. Meyer, M. Ongyerth, M. Siebauer, C. Theunert, A. Tandon, P. Moorjani, J. Pickrell, J. C. Mullikin, S. H. Vohr, R. E. Green, I. Hellmann, P. L. F. Johnson, H. Blanche, H. Cann, J. O. Kitzman, J. Shendure, E. E. Eichler, E. S. Lein, T. E. Bakken, L. V. Golovanova, V. B. Doronichev, M. V. Shunkov, A. P. Derevianko, B. Viola, M. Slatkin, D. Reich, J. Kelso, S. Pääbo, The complete genome sequence of a Neanderthal from the Altai Mountains. Nature 505, 43–49 (2014). doi:10.1038/nature12886 Medline

8. Q. Fu, C. Posth, M. Hajdinjak, M. Petr, S. Mallick, D. Fernandes, A. Furtwängler, W. Haak, M. Meyer, A. Mittnik, B. Nickel, A. Peltzer, N. Rohland, V. Slon, S. Talamo, I. Lazaridis, M. Lipson, I. Mathieson, S. Schiffels, P. Skoglund, A. P. Derevianko, N. Drozdov, V. Slavinsky, A. Tsybankov, R. G. Cremonesi, F. Mallegni, B. Gély, E. Vacca, M. R. Morales, L. G. Straus, C. Neugebauer-Maresch, M. Teschler-Nicola, S. Constantin, O. T. Moldovan, S. Benazzi, M. Peresani, D. Coppola, M. Lari, S. Ricci, A. Ronchitelli, F. Valentin, C. Thevenet, K. Wehrberger, D. Grigorescu, H. Rougier, I. Crevecoeur, D. Flas, P. Semal, M. A. Mannino, C. Cupillard, H. Bocherens, N. J. Conard, K. Harvati, V. Moiseyev, D. G. Drucker, J. Svoboda, M. P. Richards, D. Caramelli, R. Pinhasi, J. Kelso, N. Patterson, J. Krause, S. Pääbo, D. Reich, The genetic history of Ice Age Europe. Nature 534, 200–205 (2016). doi:10.1038/nature17993 Medline

9. C. Posth, G. Renaud, A. Mittnik, D. G. Drucker, H. Rougier, C. Cupillard, F. Valentin, C. Thevenet, A. Furtwängler, C. Wißing, M. Francken, M. Malina, M. Bolus, M. Lari, E. Gigli, G. Capecchi, I. Crevecoeur, C. Beauval, D. Flas, M. Germonpré, J. van der Plicht, R. Cottiaux, B. Gély, A. Ronchitelli, K. Wehrberger, D. Grigorescu, J. Svoboda, P. Semal, D. Caramelli, H. Bocherens, K. Harvati, N. J. Conard, W. Haak, A. Powell, J. Krause, Pleistocene mitochondrial genomes suggest a single major dispersal of non-Africans and a late glacial population turnover in Europe. Curr. Biol. 26, 827–833 (2016). doi:10.1016/j.cub.2016.02.022 Medline

10. J. Krause, A. W. Briggs, M. Kircher, T. Maricic, N. Zwyns, A. Derevianko, S. Pääbo, A complete mtDNA genome of an early modern human from Kostenki, Russia. Curr. Biol. 20, 231–236 (2010). doi:10.1016/j.cub.2009.11.068 Medline

11. A. Seguin-Orlando, T. S. Korneliussen, M. Sikora, A.-S. Malaspinas, A. Manica, I. Moltke, A. Albrechtsen, A. Ko, A. Margaryan, V. Moiseyev, T. Goebel, M. Westaway, D. Lambert, V. Khartanovich, J. D. Wall, P. R. Nigst, R. A. Foley, M. M. Lahr, R. Nielsen, L. Orlando, E. Willerslev, Genomic structure in Europeans dating back at least 36,200 years. Science 346, 1113–1118 (2014). doi:10.1126/science.aaa0114 Medline

12. J. Kelleher, A. M. Etheridge, G. McVean, Efficient coalescent simulation and genealogical analysis for large sample sizes. PLOS Comput. Biol. 12, e1004842 (2016). doi:10.1371/journal.pcbi.1004842 Medline

13. M. Meyer, M. Kircher, M.-T. Gansauge, H. Li, F. Racimo, S. Mallick, J. G. Schraiber, F. Jay, K. Prüfer, C. de Filippo, P. H. Sudmant, C. Alkan, Q. Fu, R. Do, N. Rohland, A. Tandon, M. Siebauer, R. E. Green, K. Bryc, A. W. Briggs, U. Stenzel, J. Dabney, J. Shendure, J. Kitzman, M. F. Hammer, M. V. Shunkov, A. P. Derevianko, N. Patterson, A. M. Andrés, E. E. Eichler, M. Slatkin, D. Reich, J. Kelso, S. Pääbo, A high-coverage genome sequence from an archaic Denisovan individual. Science 338, 222–226 (2012). doi:10.1126/science.1224344 Medline

14. M. Kirin, R. McQuillan, C. S. Franklin, H. Campbell, P. M. McKeigue, J. F. Wilson, Genomic runs of homozygosity record population history and consanguinity. PLOS ONE 5, e13996 (2010). doi:10.1371/journal.pone.0013996 Medline

15. Q. Fu, H. Li, P. Moorjani, F. Jay, S. M. Slepchenko, A. A. Bondarev, P. L. F. Johnson, A. Aximu-Petri, K. Prüfer, C. de Filippo, M. Meyer, N. Zwyns, D. C. Salazar-García, Y. V. Kuzmin, S. G. Keates, P. A. Kosintsev, D. I. Razhev, M. P. Richards, N. V. Peristov, M. Lachmann, K. Douka, T. F. G. Higham, M. Slatkin, J.-J. Hublin, D. Reich, J. Kelso, T. B. Viola, S. Pääbo, Genome sequence of a 45,000-year-old modern human from western Siberia. Nature 514, 445–449 (2014). doi:10.1038/nature13810 Medline

16. B. Vernot, J. M. Akey, Resurrecting surviving Neandertal lineages from modern human genomes. Science 343, 1017–1021 (2014). doi:10.1126/science.1245938 Medline

17. I. Juric, S. Aeschbacher, G. Coop, The strength of selection against Neanderthal introgression. PLOS Genet. 12, e1006340 (2016). doi:10.1371/journal.pgen.1006340 Medline

18. K. Harris, R. Nielsen, The genetic cost of Neanderthal introgression. Genetics 203, 881–891 (2016). doi:10.1534/genetics.116.186890 Medline

19. J. N. Fenner, Cross-cultural estimation of the human generation interval for use in genetics-based population divergence studies. Am. J. Phys. Anthropol. 128, 415–423 (2005). doi:10.1002/ajpa.20188 Medline

20. K. R. Hill, R. S. Walker, M. Bozicević, J. Eder, T. Headland, B. Hewlett, A. M. Hurtado, F. Marlowe, P. Wiessner, B. Wood, Co-residence patterns in hunter-gatherer societies show unique human social structure. Science 331, 1286–1289 (2011). doi:10.1126/science.1199071 Medline

21. M. Dyble, G. D. Salali, N. Chaudhary, A. Page, D. Smith, J. Thompson, L. Vinicius, R. Mace, A. B. Migliano, Sex equality can explain the unique social structure of hunter-gatherer bands. Science 348, 796–798 (2015). doi:10.1126/science.aaa5139 Medline

22. H. Floss, Rohmaterialversorgung im Paläolithikum des Mittelrheingebietes. Archäol. Inform. 14, 113–117 (1991). doi:10.11588/ai.1991.1.21379

23. R. L. Kelly, The Lifeways of Hunter-Gatherers: The Foraging Spectrum (Cambridge Univ. Press, 2013).

24. R. Boyd, R. H. Schonmann, R. Vicente, Hunter-gatherer population structure and the evolution of contingent cooperation. Evol. Hum. Behav. 35, 219–227 (2014). doi:10.1016/j.evolhumbehav.2014.02.002

25. A. B. Migliano, A. E. Page, J. Gómez-Gardeñes, G. D. Salali, S. Viguier, M. Dyble, J. Thompson, N. Chaudhary, D. Smith, J. Strods, R. Mace, M. G. Thomas, V. Latora, L. Vinicius, Characterization of hunter-gatherer networks and implications for cumulative culture. Nat. Hum. Behav. 1, 0043 (2017). doi:10.1038/s41562-016-0043

26. N. O. Bader, Sunghir’: An Upper Palaeolithic Site (Nauka, 1978). [In Russian]

27. T. I. Alexeeva, N. O. Bader, Eds., Homo sungirensis. Upper Palaeolithic Man: Ecological and Evolutionary Aspects of the Investigation (Scientific World, Moscow, 2000).

28. O. N. Bader, N. O. Bader, in Homo sungirensis. Upper Palaeolithic Man: Ecological and Evolutionary Aspects of the Investigation, T. I. Alexeeva, N. O. Bader, Eds. (Scientific World, Moscow, 2000), pp. 21–29.

29. P. B. Pettitt, N. O. Bader, Direct AMS radiocarbon dates for the Sungir mid Upper Palaeolithic burials. Antiquity 74, 269–270 (2000). doi:10.1017/S0003598X00059196

30. A. A. Zubov, V. M. Kharitonov, Sungir Anthropological Investigations (Scientific World, Moscow, 1984).

31. N. O. Bader, Upper Palaeolithic Site Sungir (Graves and Environment) (Scientific World, Moscow, 1998).

32. P. J. Reimer, E. Bard, A. Bayliss, J. W. Beck, P. G. Blackwell, C. B. Ramsey, C. E. Buck, H. Cheng, R. L. Edwards, M. Friedrich, P. M. Grootes, T. P. Guilderson, H. Haflidason, I. Hajdas, C. Hatté, T. J. Heaton, D. L. Hoffmann, A. G. Hogg, K. A. Hughen, K. F. Kaiser, B. Kromer, S. W. Manning, M. Niu, R. W. Reimer, D. A. Richards, E. M. Scott, J. R. Southon, R. A. Staff, C. S. M. Turney, J. van der Plicht, IntCal13 and Marine13 radiocarbon age calibration curves 0–50,000 years cal BP. Radiocarbon 55, 1869–1887 (2013). doi:10.2458/azu_js_rc.55.16947

33. C. B. Ramsey, Bayesian analysis of radiocarbon dates. Radiocarbon 51, 337–360 (2009). doi:10.1017/s0033822200033865

34. A. P. Buzhilova, The environment and health condition of the upper palaeolithic sunghir people of Russia. J. Physiol. Anthropol. Appl. Human Sci. 24, 413–418 (2005). doi:10.2114/jpa.24.413 Medline

35. E. Trinkaus, A. P. Buzhilova, The death and burial of Sunghir 1. Int. J. Osteoarchaeol. 22, 655–666 (2012). doi:10.1002/oa.1227

36. R. White, in Before Lascaux: The Complex Record of the Early Upper Paleolithic, H. Knecht, A. Pike-Tay, R. White, Eds. (CRC Press, 1993), pp. 277–299.

37. L. W. Cowgill, M. B. Mednikova, A. P. Buzhilova, E. Trinkaus, The Sunghir 3 Upper Paleolithic juvenile: Pathology versus persistence in the Paleolithic. Int. J. Osteoarchaeol. 25, 176–187 (2015). doi:10.1002/oa.2273

38. D. Guatelli-Steinberg, A. P. Buzhilova, E. Trinkaus, Developmental stress and survival among the Mid Upper Paleolithic Sunghir children: Dental enamel hypoplasias of Sunghir 2 and 3. Int. J. Osteoarchaeol. 23, 421–431 (2013). doi:10.1002/oa.1263

39. B. Pinilla, E. Trinkaus, Buccal dental microwear and diet of the Sunghir Upper Paleolithic modern humans. Archaeol. Ethnol. Anthropol. Eurasia 42, 131–142 (2014). doi:10.1016/j.aeae.2015.01.013

40. A. P. Buzhilova, in Homo sungirensis. Upper Palaeolithic Man: Ecological and Evolutionary Aspects of the Investigation, T. I. Alexeeva, N. O. Bader, Eds. (Scientifc World, Moscow, 2000), pp. 302–314.

41. V. Formicola, A. P. Buzhilova, Double child burial from Sunghir (Russia): Pathology and inferences for upper paleolithic funerary practices. Am. J. Phys. Anthropol. 124, 189–198 (2004). doi:10.1002/ajpa.10273 Medline

42. V. Formicola, From the Sunghir children to the Romito dwarf. Curr. Anthropol. 48, 446–453 (2007). doi:10.1086/517592

43. J. Riel-Salvatore, C. Gravel-Miguel, in The Oxford Handbook of the Archaeology of Death and Burial (Oxford Univ. Press, 2013), pp. 303–346.

44. P. Pettitt, The Palaeolithic Origins of Human Burial (Routledge, 2011).

45. P. Mellars, Neanderthals and the modern human colonization of Europe. Nature 432, 461–465 (2004). doi:10.1038/nature03103 Medline

46. R. G. Klein, Out of Africa and the evolution of human behavior. Evol. Anthropol. 17, 267–281 (2008). doi:10.1002/evan.20181

47. J. A. Svoboda, O. Bar-Yosef, Stránská skála. Origins of the Upper Paleolithic in the Brno Basin, Moravia, Czech Republic (Peabody Museum of Archaeology and Ethnology, Harvard University, 2003).

48. G. B. Tostevin, Seeing Lithics: A Middle-Range Theory for Testing for Cultural Transmission in the Pleistocene (Oxbow, 2012).

49. P. R. Nigst, in Living in the Landscape: Essays in Honour of Graeme Barker, K. Boyle, R. J. Rabett, C. O. Hunt, Eds. (McDonald Institute for Archaeological Research, Cambridge, 2014), pp. 35–47.

50. P. R. Nigst, P. Haesaerts, F. Damblon, C. Frank-Fellner, C. Mallol, B. Viola, M. Götzinger, L. Niven, G. Trnka, J.-J. Hublin, Early modern human settlement of Europe north of the Alps occurred 43,500 years ago in a cold steppe-type environment. Proc. Natl. Acad. Sci. U.S.A. 111, 14394–14399 (2014). doi:10.1073/pnas.1412201111 Medline

51. M. D. Bosch, M. A. Mannino, A. L. Prendergast, T. C. O’Connell, B. Demarchi, S. M. Taylor, L. Niven, J. van der Plicht, J.-J. Hublin, New chronology for Ksâr ’Akil (Lebanon) supports Levantine route of modern human dispersal into Europe. Proc. Natl. Acad. Sci. U.S.A. 112, 7683–7688 (2015). doi:10.1073/pnas.1501529112 Medline

52. J. Zilhão, Neandertals and moderns mixed, and it matters. Evol. Anthropol. 15, 183–195 (2006). doi:10.1002/evan.20110

53. N. Teyssandier, Revolution or evolution: The emergence of the Upper Paleolithic in Europe. World Archaeol. 40, 493–519 (2008). doi:10.1080/00438240802452676

54. A. A. Sinitsyn, in The Chronology of the Aurignacian and of the Transitional Technocomplexes: Dating, Stratigraphies, Cultural Implications, J. Zilhao, F. d’Errico, Eds. (Instituto Português de Arqueologia, Lisboa, 2003), pp. 89–107.

55. E. Trinkaus, O. Moldovan, S. Milota, A. Bîlgăr, L. Sarcina, S. Athreya, S. E. Bailey, R. Rodrigo, G. Mircea, T. Higham, C. B. Ramsey, J. van der Plicht, An early modern human from the Peştera cu Oase, Romania. Proc. Natl. Acad. Sci. U.S.A. 100, 11231–11236 (2003). doi:10.1073/pnas.2035108100 Medline

56. J. Svoboda, The depositional context of the early upper paleolithic human fossils from the Koneprusy (Zlatý kůn) and Mladec caves, Czech republic. J. Hum. Evol. 38, 523–536 (2000). doi:10.1006/jhev.1999.0361 Medline

57. M. White, P. Pettitt, Ancient digs and modern myths: The age and context of the Kent’s Cavern 4 maxilla and the earliest Homo sapiens specimens in Europe. Eur. J. Archaeol. 15, 392–420 (2012). doi:10.1179/1461957112Y.0000000019

58. T. Higham, T. Compton, C. Stringer, R. Jacobi, B. Shapiro, E. Trinkaus, B. Chandler, F. Gröning, C. Collins, S. Hillson, P. O’Higgins, C. FitzGerald, M. Fagan, The earliest evidence for anatomically modern humans in northwestern Europe. Nature 479, 521–524 (2011). doi:10.1038/nature10484 Medline

59. C. Proctor, K. Douka, J. W. Proctor, T. Higham, The age and context of the KC4 maxilla, Kent’s Cavern, UK. Eur. J. Archaeol. 20, 74–97 (2017). doi:10.1017/eaa.2016.1

60. S. Benazzi, K. Douka, C. Fornai, C. C. Bauer, O. Kullmer, J. Svoboda, I. Pap, F. Mallegni, P. Bayle, M. Coquerelle, S. Condemi, A. Ronchitelli, K. Harvati, G. W. Weber, Early dispersal of modern humans in Europe and implications for Neanderthal behaviour. Nature 479, 525–528 (2011). doi:10.1038/nature10617 Medline

61. W. E. Banks, F. d’Errico, J. Zilhão, Human-climate interaction during the Early Upper Paleolithic: Testing the hypothesis of an adaptive shift between the Proto-Aurignacian and the Early Aurignacian. J. Hum. Evol. 64, 39–55 (2013). doi:10.1016/j.jhevol.2012.10.001 Medline

62. A. Ronchitelli, S. Benazzi, P. Boscato, K. Douka, A. Moroni, Comments on “Human-climate interaction during the Early Upper Paleolithic: Testing the hypothesis of an adaptive shift between the Proto-Aurignacian and the Early Aurignacian” by William E. Banks, Francesco d’Errico, João Zilhão. J. Hum. Evol. 73, 107–111 (2014). doi:10.1016/j.jhevol.2013.12.010 Medline

63. J. Zilhão, W. E. Banks, F. d’Errico, P. Gioia, Analysis of site formation and assemblage integrity does not support attribution of the Uluzzian to modern humans at Grotta del Cavallo. PLOS ONE 10, e0131181 (2015). doi:10.1371/journal.pone.0131181 Medline

64. J. A. Svoboda, V. Loek, E. Vlèek, Hunters Between East and West. The Paleolithic of Moravia (Plenum, 1996).

65. E. Trinkaus, J. Svoboda, Early Modern Human Evolution in Central Europe: The People of Dolní Vstonice and Pavlov (Oxford Univ. Press, 2006).

66. A. A. Sinitsyn, in La spiritualité (Etudes et Recherches Archeologiques de l’Universite de Liege), M. Otte, Ed. (2004), pp. 237–244.

67. V. Formicola, in Hominidae: Proceedings of the 2nd International Congress of Human Paleontology (1989), pp. 483–486.

68. M. Mussi, On the chronology of the burials found in the Grimaldi Caves. Antropologia Contemporanea 9, 95–104 (1986).

69. P. B. Pettitt, M. Richards, R. Maggi, V. Formicola, The Gravettian burial known as the Prince (“Il Principe”): New evidence for his age and diet. Antiquity 77, 15–19 (2003). doi:10.1017/S0003598X00061305

70. J. Zilhão, E. Trinkaus, Portrait of the Artist as a Child. The Gravettian Human Skeleton from the Abrigo do Lagar Velho and its Archaeological Context (Instituto Português de Arqueologia, Lisboa, 2002).

71. B. Klima, Dolní Vestonice II: Ein Mammutjägerplatz und seine Bestattungen (Etudes et Recherches Archeologiques de l’Universite de Liege) (1995).

72. T. Einwögerer, H. Friesinger, M. Händel, C. Neugebauer-Maresch, U. Simon, M. Teschler-Nicola, Upper Palaeolithic infant burials. Nature 444, 285 (2006). doi:10.1038/444285a Medline

73. U. Simon, M. Händel, T. Einwögerer, C. Neugebauer-Maresch, The archaeological record of the Gravettian open air site Krems-Wachtberg. Quat. Int. 351, 5–13 (2014). doi:10.1016/j.quaint.2013.08.009

74. J. A Svoboda, The upper paleolithic burial area at Predmostí: Ritual and taphonomy. J. Hum. Evol. 54, 15–33 (2008). doi:10.1016/j.jhevol.2007.05.016 Medline

75. W. Antl-Weiser, M. Teschler-Nicola, Die menschlichen Zahnfunde von der Gravvettienfundstelle Grub/Kranawetberg bei Stillfried an der March, Niederösterreich. Archaeol. Austriaca 84–85, 201–212 (2000).

76. M. E. Teschler-Nicola, W. Antl-Weiser, H. Prossinger, Two Gravettian human deciduous teeth from Grub/Kranawetberg, lower Austria. Homo 54, 229–239 (2004). doi:10.1078/0018-442X-00074 Medline

77. C. Vercoutère, G. Giacobini, M. Patou-Mathis, Une dent humaine perforée découverte en contexte Gravettien ancien à l’abri Pataud (Dordogne, France). Anthropologie 112, 273–283 (2008). doi:10.1016/j.anthro.2008.02.002

78. P. B. Damgaard, A. Margaryan, H. Schroeder, L. Orlando, E. Willerslev, M. E. Allentoft, Improving access to endogenous DNA in ancient bones and teeth. Sci. Rep. 5, 11184 (2015). doi:10.1038/srep11184 Medline

79. J. Dabney, M. Knapp, I. Glocke, M.-T. Gansauge, A. Weihmann, B. Nickel, C. Valdiosera, N. García, S. Pääbo, J.-L. Arsuaga, M. Meyer, Complete mitochondrial genome sequence of a Middle Pleistocene cave bear reconstructed from ultrashort DNA fragments. Proc. Natl. Acad. Sci. U.S.A. 110, 15758–15763 (2013). doi:10.1073/pnas.1314445110 Medline

80. D. Y. Yang, B. Eng, J. S. Waye, J. C. Dudar, S. R. Saunders, Improved DNA extraction from ancient bones using silica-based spin columns. Am. J. Phys. Anthropol. 105, 539–543

(1998). doi:10.1002/(SICI)1096-8644(199804)105:4<539:AID-AJPA10>3.0.CO;2-1 Medline

81. C. Gamba, E. R. Jones, M. D. Teasdale, R. L. McLaughlin, G. Gonzalez-Fortes, V. Mattiangeli, L. Domboróczki, I. Kővári, I. Pap, A. Anders, A. Whittle, J. Dani, P. Raczky, T. F. G. Higham, M. Hofreiter, D. G. Bradley, R. Pinhasi, Genome flux and stasis in a five millennium transect of European prehistory. Nat. Commun. 5, 5257 (2014). doi:10.1038/ncomms6257 Medline

82. M. E. Allentoft, M. Sikora, K.-G. Sjögren, S. Rasmussen, M. Rasmussen, J. Stenderup, P. B. Damgaard, H. Schroeder, T. Ahlström, L. Vinner, A.-S. Malaspinas, A. Margaryan, T. Higham, D. Chivall, N. Lynnerup, L. Harvig, J. Baron, P. Della Casa, P. Dąbrowski, P. R. Duffy, A. V. Ebel, A. Epimakhov, K. Frei, M. Furmanek, T. Gralak, A. Gromov, S. Gronkiewicz, G. Grupe, T. Hajdu, R. Jarysz, V. Khartanovich, A. Khokhlov, V. Kiss, J. Kolář, A. Kriiska, I. Lasak, C. Longhi, G. McGlynn, A. Merkevicius, I. Merkyte, M. Metspalu, R. Mkrtchyan, V. Moiseyev, L. Paja, G. Pálfi, D. Pokutta, Ł. Pospieszny, T. D. Price, L. Saag, M. Sablin, N. Shishlina, V. Smrčka, V. I. Soenov, V. Szeverényi, G. Tóth, S. V. Trifanova, L. Varul, M. Vicze, L. Yepiskoposyan, V. Zhitenev, L. Orlando, T. Sicheritz-Pontén, S. Brunak, R. Nielsen, K. Kristiansen, E. Willerslev, Population genomics of Bronze Age Eurasia. Nature 522, 167–172 (2015). doi:10.1038/nature14507 Medline

83. M. Meyer, M. Kircher, Illumina sequencing library preparation for highly multiplexed target capture and sequencing. Cold Spring Harb. Protoc. (2010). doi:10.1101/pdb.prot5448

84. A. Seguin-Orlando, M. Schubert, J. Clary, J. Stagegaard, M. T. Alberdi, J. L. Prado, A. Prieto, E. Willerslev, L. Orlando, Ligation bias in Illumina next-generation DNA libraries: Implications for sequencing ancient genomes. PLOS ONE 8, e78575 (2013). doi:10.1371/journal.pone.0078575 Medline

85. A. W. Briggs, U. Stenzel, P. L. F. Johnson, R. E. Green, J. Kelso, K. Prüfer, M. Meyer, J. Krause, M. T. Ronan, M. Lachmann, S. Pääbo, Patterns of damage in genomic DNA sequences from a Neandertal. Proc. Natl. Acad. Sci. U.S.A. 104, 14616–14621 (2007). doi:10.1073/pnas.0704665104 Medline

86. A. W. Briggs, U. Stenzel, M. Meyer, J. Krause, M. Kircher, S. Pääbo, Removal of deaminated cytosines and detection of in vivo methylation in ancient DNA. Nucleic Acids Res. 38, e87 (2010). doi:10.1093/nar/gkp1163 Medline

87. M. Schubert, L. Ermini, C. Der Sarkissian, H. Jónsson, A. Ginolhac, R. Schaefer, M. D. Martin, R. Fernández, M. Kircher, M. McCue, E. Willerslev, L. Orlando, Characterization of ancient and modern genomes by SNP detection and phylogenomic and metagenomic analysis using PALEOMIX. Nat. Protoc. 9, 1056–1082 (2014). doi:10.1038/nprot.2014.063 Medline

88. J. T. Vilstrup, A. Seguin-Orlando, M. Stiller, A. Ginolhac, M. Raghavan, S. C. A. Nielsen, J. Weinstock, D. Froese, S. K. Vasiliev, N. D. Ovodov, J. Clary, K. M. Helgen, R. C. Fleischer, A. Cooper, B. Shapiro, L. Orlando, Mitochondrial phylogenomics of modern and ancient equids. PLOS ONE 8, e55950 (2013). doi:10.1371/journal.pone.0055950 Medline

89. M. Schubert, A. Ginolhac, S. Lindgreen, J. F. Thompson, K. A. S. Al-Rasheid, E. Willerslev, A. Krogh, L. Orlando, Improving ancient DNA read mapping against modern reference genomes. BMC Genomics 13, 178 (2012). doi:10.1186/1471-2164-13-178 Medline

90. S. Lindgreen, AdapterRemoval: Easy cleaning of next-generation sequencing reads. BMC Res. Notes 5, 337 (2012). doi:10.1186/1756-0500-5-337 Medline

91. H. Li, R. Durbin, Fast and accurate short read alignment with Burrows-Wheeler transform. Bioinformatics 25, 1754–1760 (2009). doi:10.1093/bioinformatics/btp324 Medline

92. A. McKenna, M. Hanna, E. Banks, A. Sivachenko, K. Cibulskis, A. Kernytsky, K. Garimella, D. Altshuler, S. Gabriel, M. Daly, M. A. DePristo, The Genome Analysis Toolkit: A MapReduce framework for analyzing next-generation DNA sequencing data. Genome Res. 20, 1297–1303 (2010). doi:10.1101/gr.107524.110 Medline

93. H. Jónsson, A. Ginolhac, M. Schubert, P. L. F. Johnson, L. Orlando, mapDamage2.0: Fast approximate Bayesian estimates of ancient DNA damage parameters. Bioinformatics 29, 1682–1684 (2013). doi:10.1093/bioinformatics/btt193 Medline

94. P. Skoglund, J. Storå, A. Götherström, M. Jakobsson, Accurate sex identification of ancient human remains using DNA shotgun sequencing. J. Archaeol. Sci. 40, 4477–4482 (2013). doi:10.1016/j.jas.2013.07.004

95. T. S. Korneliussen, A. Albrechtsen, R. Nielsen, ANGSD: Analysis of Next Generation Sequencing Data. BMC Bioinformatics 15, 356 (2014). doi:10.1186/s12859-014-0356-4 Medline

96. M. Rasmussen, X. Guo, Y. Wang, K. E. Lohmueller, S. Rasmussen, A. Albrechtsen, L. Skotte, S. Lindgreen, M. Metspalu, T. Jombart, T. Kivisild, W. Zhai, A. Eriksson, A. Manica, L. Orlando, F. M. De La Vega, S. Tridico, E. Metspalu, K. Nielsen, M. C. Ávila-Arcos, J. V. Moreno-Mayar, C. Muller, J. Dortch, M. T. P. Gilbert, O. Lund, A. Wesolowska, M. Karmin, L. A. Weinert, B. Wang, J. Li, S. Tai, F. Xiao, T. Hanihara, G. van Driem, A. R. Jha, F.-X. Ricaut, P. de Knijff, A. B. Migliano, I. Gallego Romero, K. Kristiansen, D. M. Lambert, S. Brunak, P. Forster, B. Brinkmann, O. Nehlich, M. Bunce, M. Richards, R. Gupta, C. D. Bustamante, A. Krogh, R. A. Foley, M. M. Lahr, F. Balloux, T. Sicheritz-Pontén, R. Villems, R. Nielsen, J. Wang, E. Willerslev, An Aboriginal Australian genome reveals separate human dispersals into Asia. Science 334, 94–98 (2011). doi:10.1126/science.1211177 Medline

97. 1000 Genomes Project Consortium, An integrated map of genetic variation from 1,092 human genomes. Nature 491, 56–65 (2012). doi:10.1038/nature11632 Medline

98. M. Stuiver, H. A. Polach, Discussion reporting of 14C data. Radiocarbon 19, 355–363 (1977). doi:10.1017/S0033822200003672

99. F. Brock, T. Higham, P. Ditchfield, C. B. Ramsey, Current pretreatment methods for AMS radiocarbon dating at the Oxford Radiocarbon Accelerator Unit (ORAU). Radiocarbon 52, 103–112 (2010). doi:10.1017/S0033822200045069

100. P. J. Reimer, M. G. L. Baillie, E. Bard, A. Bayliss, J. W. Beck, P. G. Blackwell, C. Bronk Ramsey, C. E. Buck, G. S. Burr, R. L. Edwards, M. Friedrich, P. M. Grootes, T. P. Guilderson, I. Hajdas, T. J. Heaton, A. G. Hogg, K. A. Hughen, K. F. Kaiser, B. Kromer, F. G. McCormac, S. W. Manning, R. W. Reimer, D. A. Richards, J. R. Southon, S. Talamo, C. S. M. Turney, J. van der Plicht, C. E. Weyhenmeyer, IntCal09 and Marine09 radiocarbon age calibration curves, 0–50,000 years cal BP. Radiocarbon 51, 1111–1150 (2009). doi:10.1017/S0033822200034202

101. S. Nalawade-Chavan, J. McCullagh, R. Hedges, C. Bonsall, A. Boroneanţ, C. B. Ramsey, T. Higham, Compound-specific radiocarbon dating of essential and non-essential amino acids: Towards determination of dietary reservoir effects in humans. Radiocarbon 55, 709–719 (2013). doi:10.1017/S0033822200057866

102. H. Li, B. Handsaker, A. Wysoker, T. Fennell, J. Ruan, N. Homer, G. Marth, G. Abecasis, R. Durbin, The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009). doi:10.1093/bioinformatics/btp352 Medline

103. H. Li, A statistical framework for SNP calling, mutation discovery, association mapping and population genetical parameter estimation from sequencing data. Bioinformatics 27, 2987–2993 (2011). doi:10.1093/bioinformatics/btr509 Medline

104. W. Haak, I. Lazaridis, N. Patterson, N. Rohland, S. Mallick, B. Llamas, G. Brandt, S. Nordenfelt, E. Harney, K. Stewardson, Q. Fu, A. Mittnik, E. Bánffy, C. Economou, M. Francken, S. Friederich, R. G. Pena, F. Hallgren, V. Khartanovich, A. Khokhlov, M. Kunst, P. Kuznetsov, H. Meller, O. Mochalov, V. Moiseyev, N. Nicklisch, S. L. Pichler, R. Risch, M. A. Rojo Guerra, C. Roth, A. Szécsényi-Nagy, J. Wahl, M. Meyer, J. Krause, D. Brown, D. Anthony, A. Cooper, K. W. Alt, D. Reich, Massive migration from the steppe was a source for Indo-European languages in Europe. Nature 522, 207–211 (2015). doi:10.1038/nature14317 Medline

105. Q. Fu, M. Hajdinjak, O. T. Moldovan, S. Constantin, S. Mallick, P. Skoglund, N. Patterson, N. Rohland, I. Lazaridis, B. Nickel, B. Viola, K. Prüfer, M. Meyer, J. Kelso, D. Reich, S. Pääbo, An early modern human from Romania with a recent Neanderthal ancestor. Nature 524, 216–219 (2015). doi:10.1038/nature14558 Medline

106. M. Raghavan, P. Skoglund, K. E. Graf, M. Metspalu, A. Albrechtsen, I. Moltke, S. Rasmussen, T. W. Stafford Jr., L. Orlando, E. Metspalu, M. Karmin, K. Tambets, S. Rootsi, R. Mägi, P. F. Campos, E. Balanovska, O. Balanovsky, E. Khusnutdinova, S. Litvinov, L. P. Osipova, S. A. Fedorova, M. I. Voevoda, M. DeGiorgio, T. Sicheritz-Ponten, S. Brunak, S. Demeshchenko, T. Kivisild, R. Villems, R. Nielsen, M. Jakobsson, E. Willerslev, Upper Palaeolithic Siberian genome reveals dual ancestry of Native Americans. Nature 505, 87–91 (2014). doi:10.1038/nature12736 Medline

107. P. Skoglund, H. Malmström, A. Omrak, M. Raghavan, C. Valdiosera, T. Günther, P. Hall, K. Tambets, J. Parik, K.-G. Sjögren, J. Apel, E. Willerslev, J. Storå, A. Götherström, M. Jakobsson, Genomic diversity and admixture differs for Stone-Age Scandinavian foragers and farmers. Science 344, 747–750 (2014). doi:10.1126/science.1253448 Medline

108. E. R. Jones, G. Gonzalez-Fortes, S. Connell, V. Siska, A. Eriksson, R. Martiniano, R. L. McLaughlin, M. Gallego Llorente, L. M. Cassidy, C. Gamba, T. Meshveliani, O. Bar-Yosef, W. Müller, A. Belfer-Cohen, Z. Matskevich, N. Jakeli, T. F. G. Higham, M. Currat, D. Lordkipanidze, M. Hofreiter, A. Manica, R. Pinhasi, D. G. Bradley, Upper Palaeolithic genomes reveal deep roots of modern Eurasians. Nat. Commun. 6, 8912 (2015). doi:10.1038/ncomms9912 Medline

109. I. Mathieson, I. Lazaridis, N. Rohland, S. Mallick, N. Patterson, S. A. Roodenberg, E. Harney, K. Stewardson, D. Fernandes, M. Novak, K. Sirak, C. Gamba, E. R. Jones, B. Llamas, S. Dryomov, J. Pickrell, J. L. Arsuaga, J. M. B. de Castro, E. Carbonell, F. Gerritsen, A. Khokhlov, P. Kuznetsov, M. Lozano, H. Meller, O. Mochalov, V. Moiseyev, M. A. R. Guerra, J. Roodenberg, J. M. Vergès, J. Krause, A. Cooper, K. W. Alt, D. Brown, D. Anthony, C. Lalueza-Fox, W. Haak, R. Pinhasi, D. Reich, Genome-wide patterns of selection in 230 ancient Eurasians. Nature 528, 499–503 (2015). doi:10.1038/nature16152 Medline

110. I. Olalde, H. Schroeder, M. Sandoval-Velasco, L. Vinner, I. Lobón, O. Ramirez, S. Civit, P. García Borja, D. C. Salazar-García, S. Talamo, J. María Fullola, F. Xavier Oms, M. Pedro, P. Martínez, M. Sanz, J. Daura, J. Zilhão, T. Marquès-Bonet, M. T. Gilbert, C. Lalueza-Fox, A common genetic origin for early farmers from Mediterranean Cardial and Central European LBK cultures. Mol. Biol. Evol. 32, 3132–3142 (2015). doi:10.1093/molbev/msv181 Medline

111. I. Lazaridis, D. Nadel, G. Rollefson, D. C. Merrett, N. Rohland, S. Mallick, D. Fernandes, M. Novak, B. Gamarra, K. Sirak, S. Connell, K. Stewardson, E. Harney, Q. Fu, G. Gonzalez-Fortes, E. R. Jones, S. A. Roodenberg, G. Lengyel, F. Bocquentin, B. Gasparian, J. M. Monge, M. Gregg, V. Eshed, A.-S. Mizrahi, C. Meiklejohn, F. Gerritsen, L. Bejenaru, M. Blüher, A. Campbell, G. Cavalleri, D. Comas, P. Froguel, E. Gilbert, S. M. Kerr, P. Kovacs, J. Krause, D. McGettigan, M. Merrigan, D. A. Merriwether, S. O’Reilly, M. B. Richards, O. Semino, M. Shamoon-Pour, G. Stefanescu, M. Stumvoll, A. Tönjes, A. Torroni, J. F. Wilson, L. Yengo, N. A. Hovhannisyan, N.

Patterson, R. Pinhasi, D. Reich, Genomic insights into the origin of farming in the ancient Near East. Nature 536, 419–424 (2016). doi:10.1038/nature19310 Medline

112. F. Broushaki, M. G. Thomas, V. Link, S. López, L. van Dorp, K. Kirsanow, Z. Hofmanová, Y. Diekmann, L. M. Cassidy, D. Díez-Del-Molino, A. Kousathanas, C. Sell, H. K. Robson, R. Martiniano, J. Blöcher, A. Scheu, S. Kreutzer, R. Bollongino, D. Bobo, H. Davudi, O. Munoz, M. Currat, K. Abdi, F. Biglari, O. E. Craig, D. G. Bradley, S. Shennan, K. Veeramah, M. Mashkour, D. Wegmann, G. Hellenthal, J. Burger, Early Neolithic genomes from the eastern Fertile Crescent. Science 353, 499–503 (2016). doi:10.1126/science.aaf7943 Medline

113. Z. Hofmanová, S. Kreutzer, G. Hellenthal, C. Sell, Y. Diekmann, D. Díez-Del-Molino, L. van Dorp, S. López, A. Kousathanas, V. Link, K. Kirsanow, L. M. Cassidy, R. Martiniano, M. Strobel, A. Scheu, K. Kotsakis, P. Halstead, S. Triantaphyllou, N. Kyparissi-Apostolika, D. Urem-Kotsou, C. Ziota, F. Adaktylou, S. Gopalan, D. M. Bobo, L. Winkelbach, J. Blöcher, M. Unterländer, C. Leuenberger, Ç. Çilingiroğlu, B. Horejs, F. Gerritsen, S. J. Shennan, D. G. Bradley, M. Currat, K. R. Veeramah, D. Wegmann, M. G. Thomas, C. Papageorgopoulou, J. Burger, Early farmers from across Europe directly descended from Neolithic Aegeans. Proc. Natl. Acad. Sci. U.S.A. 113, 6886–6891 (2016). doi:10.1073/pnas.1523951113 Medline

114. G. M. Kılınç, A. Omrak, F. Özer, T. Günther, A. M. Büyükkarakaya, E. Bıçakçı, D. Baird, H. M. Dönertaş, A. Ghalichi, R. Yaka, D. Koptekin, S. C. Açan, P. Parvizi, M. Krzewińska, E. A. Daskalaki, E. Yüncü, N. D. Dağtaş, A. Fairbairn, J. Pearson, G. Mustafaoğlu, Y. S. Erdal, Y. G. Çakan, İ. Togan, M. Somel, J. Storå, M. Jakobsson, A. Götherström, The demographic development of the first farmers in Anatolia. Curr. Biol. 26, 2659–2666 (2016). doi:10.1016/j.cub.2016.07.057 Medline

115. S. Mallick, H. Li, M. Lipson, I. Mathieson, M. Gymrek, F. Racimo, M. Zhao, N. Chennagiri, S. Nordenfelt, A. Tandon, P. Skoglund, I. Lazaridis, S. Sankararaman, Q. Fu, N. Rohland, G. Renaud, Y. Erlich, T. Willems, C. Gallo, J. P. Spence, Y. S. Song, G. Poletti, F. Balloux, G. van Driem, P. de Knijff, I. G. Romero, A. R. Jha, D. M. Behar, C. M. Bravi, C. Capelli, T. Hervig, A. Moreno-Estrada, O. L. Posukh, E. Balanovska, O. Balanovsky, S. Karachanak-Yankova, H. Sahakyan, D. Toncheva, L. Yepiskoposyan, C. Tyler-Smith, Y. Xue, M. S. Abdullah, A. Ruiz-Linares, C. M. Beall, A. Di Rienzo, C. Jeong, E. B. Starikovskaya, E. Metspalu, J. Parik, R. Villems, B. M. Henn, U. Hodoglugil, R. Mahley, A. Sajantila, G. Stamatoyannopoulos, J. T. S. Wee, R. Khusainova, E. Khusnutdinova, S. Litvinov, G. Ayodo, D. Comas, M. F. Hammer, T. Kivisild, W. Klitz, C. A. Winkler, D. Labuda, M. Bamshad, L. B. Jorde, S. A. Tishkoff, W. S. Watkins, M. Metspalu, S. Dryomov, R. Sukernik, L. Singh, K. Thangaraj, S. Pääbo, J. Kelso, N. Patterson, D. Reich, The Simons Genome Diversity Project: 300 genomes from 142 diverse populations. Nature 538, 201–206 (2016). doi:10.1038/nature18964 Medline

116. A.-S. Malaspinas, M. C. Westaway, C. Muller, V. C. Sousa, O. Lao, I. Alves, A. Bergström, G. Athanasiadis, J. Y. Cheng, J. E. Crawford, T. H. Heupink, E. Macholdt, S. Peischl, S. Rasmussen, S. Schiffels, S. Subramanian, J. L. Wright, A. Albrechtsen, C. Barbieri, I. Dupanloup, A. Eriksson, A. Margaryan, I. Moltke, I. Pugach, T. S. Korneliussen, I. P. Levkivskyi, J. V. Moreno-Mayar, S. Ni, F. Racimo, M. Sikora, Y. Xue, F. A. Aghakhanian, N. Brucato, S. Brunak, P. F. Campos, W. Clark, S. Ellingvåg, G. Fourmile, P. Gerbault, D. Injie, G. Koki, M. Leavesley, B. Logan, A. Lynch, E. A. Matisoo-Smith, P. J. McAllister, A. J. Mentzer, M. Metspalu, A. B. Migliano, L. Murgha, M. E. Phipps, W. Pomat, D. Reynolds, F.-X. Ricaut, P. Siba, M. G. Thomas, T. Wales, C. M. Wall, S. J. Oppenheimer, C. Tyler-Smith, R. Durbin, J. Dortch, A. Manica, M. H. Schierup, R. A. Foley, M. M. Lahr, C. Bowern, J. D. Wall, T. Mailund, M. Stoneking, R. Nielsen, M. S. Sandhu, L. Excoffier, D. M. Lambert, E. Willerslev, A genomic history of Aboriginal Australia. Nature 538, 207–214 (2016). doi:10.1038/nature18299 Medline

117. M. Mondal, F. Casals, T. Xu, G. M. Dall’Olio, M. Pybus, M. G. Netea, D. Comas, H. Laayouni, Q. Li, P. P. Majumder, J. Bertranpetit, Genomic analysis of Andamanese provides insights into ancient human migration into Asia and adaptation. Nat. Genet. 48, 1066–1070 (2016). doi:10.1038/ng.3621 Medline

118. M. Raghavan, M. DeGiorgio, A. Albrechtsen, I. Moltke, P. Skoglund, T. S. Korneliussen, B. Grønnow, M. Appelt, H. C. Gulløv, T. M. Friesen, W. Fitzhugh, H. Malmström, S. Rasmussen, J. Olsen, L. Melchior, B. T. Fuller, S. M. Fahrni, T. Stafford Jr., V. Grimes, M. A. P. Renouf, J. Cybulski, N. Lynnerup, M. M. Lahr, K. Britton, R. Knecht, J. Arneborg, M. Metspalu, O. E. Cornejo, A.-S. Malaspinas, Y. Wang, M. Rasmussen, V. Raghavan, T. V. O. Hansen, E. Khusnutdinova, T. Pierre, K. Dneprovsky, C. Andreasen, H. Lange, M. G. Hayes, J. Coltrain, V. A. Spitsyn, A. Götherström, L. Orlando, T. Kivisild, R. Villems, M. H. Crawford, F. C. Nielsen, J. Dissing, J. Heinemeier, M. Meldgaard, C. Bustamante, D. H. O’Rourke, M. Jakobsson, M. T. P. Gilbert, R. Nielsen, E. Willerslev, The genetic prehistory of the New World Arctic. Science 345, 1255832 (2014). doi:10.1126/science.1255832 Medline

119. M. Raghavan, M. Steinrücken, K. Harris, S. Schiffels, S. Rasmussen, M. DeGiorgio, A. Albrechtsen, C. Valdiosera, M. C. Ávila-Arcos, A.-S. Malaspinas, A. Eriksson, I. Moltke, M. Metspalu, J. R. Homburger, J. Wall, O. E. Cornejo, J. V. Moreno-Mayar, T. S. Korneliussen, T. Pierre, M. Rasmussen, P. F. Campos, P. de Barros Damgaard, M. E. Allentoft, J. Lindo, E. Metspalu, R. Rodríguez-Varela, J. Mansilla, C. Henrickson, A. Seguin-Orlando, H. Malmström, T. Stafford Jr., S. S. Shringarpure, A. Moreno-Estrada, M. Karmin, K. Tambets, A. Bergström, Y. Xue, V. Warmuth, A. D. Friend, J. Singarayer, P. Valdes, F. Balloux, I. Leboreiro, J. L. Vera, H. Rangel-Villalobos, D. Pettener, D. Luiselli, L. G. Davis, E. Heyer, C. P. E. Zollikofer, M. S. Ponce de León, C. I. Smith, V. Grimes, K.-A. Pike, M. Deal, B. T. Fuller, B. Arriaza, V. Standen, M. F. Luz, F. Ricaut, N. Guidon, L. Osipova, M. I. Voevoda, O. L. Posukh, O. Balanovsky, M. Lavryashina,

Y. Bogunov, E. Khusnutdinova, M. Gubina, E. Balanovska, S. Fedorova, S. Litvinov, B. Malyarchuk, M. Derenko, M. J. Mosher, D. Archer, J. Cybulski, B. Petzelt, J. Mitchell, R. Worl, P. J. Norman, P. Parham, B. M. Kemp, T. Kivisild, C. Tyler-Smith, M. S. Sandhu, M. Crawford, R. Villems, D. G. Smith, M. R. Waters, T. Goebel, J. R. Johnson, R. S. Malhi, M. Jakobsson, D. J. Meltzer, A. Manica, R. Durbin, C. D. Bustamante, Y. S. Song, R. Nielsen, E. Willerslev, Genomic evidence for the Pleistocene and recent population history of Native Americans. Science 349, aab3884 (2015). doi:10.1126/science.aab3884 Medline

120. I. Lazaridis, N. Patterson, A. Mittnik, G. Renaud, S. Mallick, K. Kirsanow, P. H. Sudmant, J. G. Schraiber, S. Castellano, M. Lipson, B. Berger, C. Economou, R. Bollongino, Q. Fu, K. I. Bos, S. Nordenfelt, H. Li, C. de Filippo, K. Prüfer, S. Sawyer, C. Posth, W. Haak, F. Hallgren, E. Fornander, N. Rohland, D. Delsate, M. Francken, J.-M. Guinet, J. Wahl, G. Ayodo, H. A. Babiker, G. Bailliet, E. Balanovska, O. Balanovsky, R. Barrantes, G. Bedoya, H. Ben-Ami, J. Bene, F. Berrada, C. M. Bravi, F. Brisighelli, G. B. J. Busby, F. Cali, M. Churnosov, D. E. C. Cole, D. Corach, L. Damba, G. van Driem, S. Dryomov, J.-M. Dugoujon, S. A. Fedorova, I. Gallego Romero, M. Gubina, M. Hammer, B. M. Henn, T. Hervig, U. Hodoglugil, A. R. Jha, S. Karachanak-Yankova, R. Khusainova, E. Khusnutdinova, R. Kittles, T. Kivisild, W. Klitz, V. Kučinskas, A. Kushniarevich, L. Laredj, S. Litvinov, T. Loukidis, R. W. Mahley, B. Melegh, E. Metspalu, J. Molina, J. Mountain, K. Näkkäläjärvi, D. Nesheva, T. Nyambo, L. Osipova, J. Parik, F. Platonov, O. Posukh, V. Romano, F. Rothhammer, I. Rudan, R. Ruizbakiev, H. Sahakyan, A. Sajantila, A. Salas, E. B. Starikovskaya, A. Tarekegn, D. Toncheva, S. Turdikulova, I. Uktveryte, O. Utevska, R. Vasquez, M. Villena, M. Voevoda, C. A. Winkler, L. Yepiskoposyan, P. Zalloua, T. Zemunik, A. Cooper, C. Capelli, M. G. Thomas, A. Ruiz-Linares, S. A. Tishkoff, L. Singh, K. Thangaraj, R. Villems, D. Comas, R. Sukernik, M. Metspalu, M. Meyer, E. E. Eichler, J. Burger, M. Slatkin, S. Pääbo, J. Kelso, D. Reich, J. Krause, Ancient human genomes suggest three ancestral populations for present-day Europeans. Nature 513, 409–413 (2014). doi:10.1038/nature13673 Medline

121. R. E. Green, J. Krause, A. W. Briggs, T. Maricic, U. Stenzel, M. Kircher, N. Patterson, H. Li, W. Zhai, M. H. Y. Fritz, N. F. Hansen, E. Y. Durand, A. S. Malaspinas, J. D. Jensen, T. Marques-Bonet, C. Alkan, K. Prüfer, M. Meyer, H. A. Burbano, J. M. Good, R. Schultz, A. Aximu-Petri, A. Butthof, B. Höber, B. Höffner, M. Siegemund, A. Weihmann, C. Nusbaum, E. S. Lander, C. Russ, N. Novod, J. Affourtit, M. Egholm, C. Verna, P. Rudan, D. Brajkovic, Ž. Kucan, I. Gušic, V. B. Doronichev, L. V. Golovanova, C. Lalueza-Fox, M. de la Rasilla, J. Fortea, A. Rosas, R. W. Schmitz, P. L. F. Johnson, E. E. Eichler, D. Falush, E. Birney, J. C. Mullikin, M. Slatkin, R. Nielsen, J. Kelso, M. Lachmann, D. Reich, S. Pääbo, A draft sequence of the Neandertal genome. Science 328, 710–722 (2010). doi:10.1126/science.1188021 Medline

122. G. Renaud, V. Slon, A. T. Duggan, J. Kelso, Schmutzi: Estimation of contamination and endogenous mitochondrial consensus calling for ancient DNA. Genome Biol. 16, 224 (2015). doi:10.1186/s13059-015-0776-0 Medline

123. H. Weissensteiner, D. Pacher, A. Kloss-Brandstätter, L. Forer, G. Specht, H.-J. Bandelt, F. Kronenberg, A. Salas, S. Schönherr, HaploGrep 2: Mitochondrial haplogroup classification in the era of high-throughput sequencing. Nucleic Acids Res. 44, W58–W63 (2016). doi:10.1093/nar/gkw233 Medline

124. A. Löytynoja, N. Goldman, Phylogeny-aware gap placement prevents errors in sequence alignment and evolutionary analysis. Science 320, 1632–1635 (2008). doi:10.1126/science.1158395 Medline

125. A. Stamatakis, RAxML version 8: A tool for phylogenetic analysis and post-analysis of large phylogenies. Bioinformatics 30, 1312–1313 (2014). doi:10.1093/bioinformatics/btu033 Medline

126. I. Olalde, M. E. Allentoft, F. Sánchez-Quinto, G. Santpere, C. W. K. Chiang, M. DeGiorgio, J. Prado-Martinez, J. A. Rodríguez, S. Rasmussen, J. Quilez, O. Ramírez, U. M. Marigorta, M. Fernández-Callejo, M. E. Prada, J. M. V. Encinas, R. Nielsen, M. G. Netea, J. Novembre, R. A. Sturm, P. Sabeti, T. Marquès-Bonet, A. Navarro, E. Willerslev, C. Lalueza-Fox, Derived immune and ancestral pigmentation alleles in a 7,000-year-old Mesolithic European. Nature 507, 225–228 (2014). doi:10.1038/nature12960 Medline

127. P. Hallast, C. Batini, D. Zadik, P. Maisano Delser, J. H. Wetton, E. Arroyo-Pardo, G. L. Cavalleri, P. de Knijff, G. Destro Bisol, B. M. Dupuy, H. A. Eriksen, L. B. Jorde, T. E. King, M. H. Larmuseau, A. López de Munain, A. M. López-Parra, A. Loutradis, J. Milasin, A. Novelletto, H. Pamjav, A. Sajantila, W. Schempp, M. Sears, A. Tolun, C. Tyler-Smith, A. Van Geystelen, S. Watkins, B. Winney, M. A. Jobling, The Y-chromosome tree bursts into leaf: 13,000 high-confidence SNPs covering the majority of known clades. Mol. Biol. Evol. 32, 661–673 (2015). doi:10.1093/molbev/msu327 Medline

128. K. Tamura, G. Stecher, D. Peterson, A. Filipski, S. Kumar, MEGA6: Molecular Evolutionary Genetics Analysis version 6.0. Mol. Biol. Evol. 30, 2725–2729 (2013). doi:10.1093/molbev/mst197 Medline

129. T. S. Korneliussen, I. Moltke, NgsRelate: A software tool for estimating pairwise relatedness from next-generation sequencing data. Bioinformatics 31, 4009–4011 (2015). doi:10.1093/bioinformatics/btv509 Medline

130. A. Albrechtsen, T. Sand Korneliussen, I. Moltke, T. van Overseem Hansen, F. C. Nielsen, R. Nielsen, Relatedness mapping and tracts of relatedness for genome-wide data in the

presence of linkage disequilibrium. Genet. Epidemiol. 33, 266–274 (2009). doi:10.1002/gepi.20378 Medline

131. A. Manichaikul, J. C. Mychaleckyj, S. S. Rich, K. Daly, M. Sale, W.-M. Chen, Robust relationship inference in genome-wide association studies. Bioinformatics 26, 2867–2873 (2010). doi:10.1093/bioinformatics/btq559 Medline

132. A. Seguin-Orlando, C. Gamba, C. Der Sarkissian, L. Ermini, G. Louvel, E. Boulygina, A. Sokolov, A. Nedoluzhko, E. D. Lorenzen, P. Lopez, H. G. McDonald, E. Scott, A. Tikhonov, T. W. Stafford Jr., A. H. Alfarhan, S. A. Alquraishi, K. A. S. Al-Rasheid, B. Shapiro, E. Willerslev, E. Prokhortchouk, L. Orlando, Pros and cons of methylation-based enrichment methods for ancient DNA. Sci. Rep. 5, 11826 (2015). doi:10.1038/srep11826 Medline

133. A. Ko, R. Nielsen, Composite likelihood method for inferring local pedigrees. PLOS Genet. 13, e1006963 (2017). doi:10.1371/journal.pgen.1006963

134. S. Purcell, B. Neale, K. Todd-Brown, L. Thomas, M. A. R. Ferreira, D. Bender, J. Maller, P. Sklar, P. I. W. de Bakker, M. J. Daly, P. C. Sham, PLINK: A tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81, 559–575 (2007). doi:10.1086/519795 Medline

135. 1000 Genomes Project Consortium, A global reference for human genetic variation. Nature 526, 68–74 (2015). doi:10.1038/nature15393 Medline

136. J. M. M. Kuhn, M. Jakobsson, T. Günther, Estimating genetic kin relationships in prehistoric populations. bioRxiv:100297 (2017).

137. International HapMap 3 Consortium, Integrating common and rare genetic variation in diverse human populations. Nature 467, 52–58 (2010). doi:10.1038/nature09298 Medline

138. A. Scally, R. Durbin, Revising the human mutation rate: Implications for understanding human evolution. Nat. Rev. Genet. 13, 745–753 (2012). doi:10.1038/nrg3295 Medline

139. B. L. Browning, S. R. Browning, Detecting identity by descent and estimating genotype error rates in sequence data. Am. J. Hum. Genet. 93, 840–851 (2013). doi:10.1016/j.ajhg.2013.09.014 Medline

140. P. F. Palamara, T. Lencz, A. Darvasi, I. Pe’er, Length distributions of identity by descent reveal fine-scale demographic history. Am. J. Hum. Genet. 91, 809–822 (2012). doi:10.1016/j.ajhg.2012.08.030 Medline

141. N. Patterson, P. Moorjani, Y. Luo, S. Mallick, N. Rohland, Y. Zhan, T. Genschoreck, T. Webster, D. Reich, Ancient admixture in human history. Genetics 192, 1065–1093 (2012). doi:10.1534/genetics.112.145037 Medline

142. N. Patterson, A. L. Price, D. Reich, Population structure and eigenanalysis. PLOS Genet. 2, e190 (2006). doi:10.1371/journal.pgen.0020190 Medline

143. R Project for Statistical Computing; www.r-project.org/.

144. M. Rasmussen, S. L. Anzick, M. R. Waters, P. Skoglund, M. DeGiorgio, T. W. Stafford Jr., S. Rasmussen, I. Moltke, A. Albrechtsen, S. M. Doyle, G. D. Poznik, V. Gudmundsdottir, R. Yadav, A.-S. Malaspinas, S. S. White 5th, M. E. Allentoft, O. E. Cornejo, K. Tambets, A. Eriksson, P. D. Heintzman, M. Karmin, T. S. Korneliussen, D. J. Meltzer, T. L. Pierre, J. Stenderup, L. Saag, V. M. Warmuth, M. C. Lopes, R. S. Malhi, S. Brunak, T. Sicheritz-Ponten, I. Barnes, M. Collins, L. Orlando, F. Balloux, A. Manica, R. Gupta, M. Metspalu, C. D. Bustamante, M. Jakobsson, R. Nielsen, E. Willerslev, The genome of a Late Pleistocene human from a Clovis burial site in western Montana. Nature 506, 225–229 (2014). doi:10.1038/nature13025 Medline

145. R. Nielsen, T. Korneliussen, A. Albrechtsen, Y. Li, J. Wang, SNP calling, genotype calling, and sample allele frequency estimation from new-generation sequencing data. PLOS ONE 7, e37558 (2012). doi:10.1371/journal.pone.0037558 Medline

146. J. C. Nash, R. Varadhan, Unifying optimization algorithms to aid software system users: optimx for R. J. Stat. Softw. 43, www.jstatsoft.org/article/view/v043i09/ (2011).

147. A. M. Adams, R. R. Hudson, Maximum-likelihood estimation of demographic parameters using the frequency spectrum of unlinked single-nucleotide polymorphisms. Genetics 168, 1699–1712 (2004). doi:10.1534/genetics.104.030171 Medline

148. R. Nielsen, Estimation of population parameters and recombination rates from single nucleotide polymorphisms. Genetics 154, 931–942 (2000). Medline

149. L. Excoffier, I. Dupanloup, E. Huerta-Sánchez, V. C. Sousa, M. Foll, Robust demographic inference from genomic and SNP data. PLOS Genet. 9, e1003905 (2013). doi:10.1371/journal.pgen.1003905 Medline

150. X.-L. Meng, D. B. Rubin, Maximum likelihood estimation via the ECM algorithm: A general framework. Biometrika 80, 267–278 (1993). doi:10.1093/biomet/80.2.267

151. R. Brent, Algorithms for Minimization Without Derivatives (Prentice-Hall, 1973).

152. F. Cunningham, M. R. Amode, D. Barrell, K. Beal, K. Billis, S. Brent, D. Carvalho-Silva, P. Clapham, G. Coates, S. Fitzgerald, L. Gil, C. G. Girón, L. Gordon, T. Hourlier, S. E. Hunt, S. H. Janacek, N. Johnson, T. Juettemann, A. K. Kähäri, S. Keenan, F. J. Martin, T. Maurel, W. McLaren, D. N. Murphy, R. Nag, B. Overduin, A. Parker, M. Patricio, E. Perry, M. Pignatelli, H. S. Riat, D. Sheppard, K. Taylor, A. Thormann, A. Vullo, S. P. Wilder, A. Zadissa, B. L. Aken, E. Birney, J. Harrow, R. Kinsella, M. Muffato, M. Ruffier, S. M. J. Searle, G. Spudich, S. J. Trevanion, A. Yates, D. R. Zerbino, P. Flicek, Ensembl 2015. Nucleic Acids Res. 43, D662–D669 (2015). doi:10.1093/nar/gku1010 Medline

153. K. R. Rosenbloom, J. Armstrong, G. P. Barber, J. Casper, H. Clawson, M. Diekhans, T. R. Dreszer, P. A. Fujita, L. Guruvadoo, M. Haeussler, R. A. Harte, S. Heitner, G. Hickey, A.

S. Hinrichs, R. Hubley, D. Karolchik, K. Learned, B. T. Lee, C. H. Li, K. H. Miga, N. Nguyen, B. Paten, B. J. Raney, A. F. A. Smit, M. L. Speir, A. S. Zweig, D. Haussler, R. M. Kuhn, W. J. Kent, The UCSC Genome Browser database: 2015 update. Nucleic Acids Res. 43, D670–D681 (2015). doi:10.1093/nar/gku1177 Medline

154. P. Flicek, M. R. Amode, D. Barrell, K. Beal, S. Brent, Y. Chen, P. Clapham, G. Coates, S. Fairley, S. Fitzgerald, L. Gordon, M. Hendrix, T. Hourlier, N. Johnson, A. Kähäri, D. Keefe, S. Keenan, R. Kinsella, F. Kokocinski, E. Kulesha, P. Larsson, I. Longden, W. McLaren, B. Overduin, B. Pritchard, H. S. Riat, D. Rios, G. R. S. Ritchie, M. Ruffier, M. Schuster, D. Sobral, G. Spudich, Y. A. Tang, S. Trevanion, J. Vandrovcova, A. J. Vilella, S. White, S. P. Wilder, A. Zadissa, J. Zamora, B. L. Aken, E. Birney, F. Cunningham, I. Dunham, R. Durbin, X. M. Fernández-Suarez, J. Herrero, T. J. P. Hubbard, A. Parker, G. Proctor, J. Vogel, S. M. J. Searle, Ensembl 2011. Nucleic Acids Res. 39, D800–D806 (2011). doi:10.1093/nar/gkq1064 Medline

155. R. N. Gutenkunst, R. D. Hernandez, S. H. Williamson, C. D. Bustamante, Inferring the joint demographic history of multiple populations from multidimensional SNP frequency data. PLOS Genet. 5, e1000695 (2009). doi:10.1371/journal.pgen.1000695 Medline

156. A. C. Davison, D. V. Hinkley, Bootstrap Methods and Their Application (Cambridge Univ. Press, 2013).

157. P. W. Messer, SLiM: Simulating evolution with selection and linkage. Genetics 194, 1037–1039 (2013). doi:10.1534/genetics.113.152181 Medline

158. A. Marom, J. S. O. McCullagh, T. F. G. Higham, R. E. M. Hedges, Hydroxyproline dating: Experiments on the 14C analysis of contaminated and low-collagen bones. Radiocarbon 55, 698–708 (2013). doi:10.1017/S0033822200057854

159. T. B. Coplen, Reporting of stable hydrogen, carbon, and oxygen isotopic abundances. Geothermics 24, 707–712 (1995). doi:10.1016/0375-6505(95)00024-0