alexandros stamatakis and alexey kozlov · 2021. 1. 13. · raxml-ng: a fast, scalable and...
TRANSCRIPT
![Page 1: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/1.jpg)
RAxML-NG: a fast, scalable and user-friendly tool for
maximum likelihood phylogenetic inference
Computational Molecular Evolution Group, Heidelberg Institute for Theoretical Studies
Alexandros Stamatakis and Alexey Kozlov
Institute for Theoretical Informatics, Karlsruhe Institute of Technology
ISCBacademy webinar – September 30, 2020
![Page 2: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/2.jpg)
Outline● Introduction to Phylogenetic Inference -
Alexandros● The RAxML Search Algorithm - Alexandros● Improvements in RAxML Next Generation -
Alexey● Tutorial - Alexey
![Page 3: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/3.jpg)
A phylogeny
Taxon 5Taxon 3
Taxon 4
Taxon 1
Taxon 2
Phylogenies describe evolutionary relationships among species
Hypothetical common ancestors
![Page 4: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/4.jpg)
A phylogeny
Taxon 5Taxon 3
Taxon 4
Taxon 1
Taxon 2 Extant species
![Page 5: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/5.jpg)
A phylogeny
Taxon 5Taxon 3
Taxon 4
Taxon 1
Taxon 2
Phylogenetic trees are unrooted binary trees!!!
![Page 6: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/6.jpg)
Tree Inference Pipeline
Taxon 1:ACGTTTTaxon 2:ACGTTTaxon 3:ACCCTTaxon 4:AGGGTTT
MSAProgram
Tree inferenceprogram
Taxon 1:ACGTTT-Taxon 2:ACGTT--Taxon 3:ACCCT--Taxon 4:AGGGTTT
Taxon 1
Taxon 2
Taxon 3
Taxon 4
![Page 7: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/7.jpg)
Tree Inference Pipeline
Taxon 1:ACGTTTTaxon 2:ACGTTTaxon 3:ACCCTTaxon 4:AGGGTTT
MSAProgram
RAxML-NGTaxon 1:ACGTTT-Taxon 2:ACGTT--Taxon 3:ACCCT--Taxon 4:AGGGTTT
Taxon 1
Taxon 2
Taxon 3
Taxon 4Our focus in this talk
![Page 8: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/8.jpg)
How many unrooted 4-taxon trees exist?
![Page 9: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/9.jpg)
How many unrooted 4-taxon trees exist?
A
D
B
C
A
C
B
D
A
B
C
D
![Page 10: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/10.jpg)
How do we select among them ?
A
D
B
C
A
C
B
D
A
B
C
D
![Page 11: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/11.jpg)
How do we select among them???
A
D
B
C
A
C
B
D
A
B
C
D
We need scoring criteria!!!
Score: 1.0
Score: 2.0Score: 3.0
![Page 12: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/12.jpg)
How do we select among them???
A
D
B
C
A
C
B
D
A
B
C
D
We need scoring criteria!!!The currently most widely used criterion is maximum likelihood:How likely is it that the tree, given a model of evolution, generatedthe observed data?
Score: 1.0
Score: 2.0Score: 3.0
![Page 13: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/13.jpg)
How do we select among them???
A
D
B
C
A
C
B
D
A
B
C
D
We need scoring criteria!!!The currently most widely used criterion is maximum likelihood:How likely is it that the tree, given a model of evolution, generatedthe observed data?
Score: 1.0
Score: 2.0Score: 3.0A
Maximum Likelihood tree
![Page 14: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/14.jpg)
14
The number of trees
3 taxa → 1 tree
![Page 15: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/15.jpg)
15
The number of trees
4 taxa → 3 trees
![Page 16: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/16.jpg)
16
The number of trees
5 taxa → 15 trees
![Page 17: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/17.jpg)
17
The number of trees
6 taxa → 105 trees
![Page 18: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/18.jpg)
18
The number of trees explodes!
BANG !
![Page 19: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/19.jpg)
19
# possible trees with 2000 taxa
![Page 20: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/20.jpg)
20
Problem Complexity
good scores
bad scores
search spaceheuristic tree search strategy
Global maximum
![Page 21: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/21.jpg)
21
Problem Complexity
good scores
bad scores
search spaceheuristic tree search strategy
Finding the best tree under Maximum Likelihood is NP-hard!
Global maximum
![Page 22: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/22.jpg)
22
Problem Complexity
search spaceheuristic tree search strategy
Maximum Likelihood tree searches can end up in local optima
Global maximum
![Page 23: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/23.jpg)
23
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
![Page 24: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/24.jpg)
24
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
![Page 25: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/25.jpg)
25
Nucleotide Substitution Models
A C
G T
We model evolution as time-reversible Markov Process!
![Page 26: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/26.jpg)
26
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Commonly denoted as Q matrix: transition probs for time dt, for time t: P(t)=eQt
![Page 27: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/27.jpg)
27
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
![Page 28: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/28.jpg)
28
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
Seq 1Seq 1
Seq 2Seq 2 Seq 4Seq 4
Seq 3Seq 3b1b1
b2b2
b5b5
b3b3
b4b4
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
![Page 29: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/29.jpg)
29
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
Seq 1Seq 1
Seq 2Seq 2 Seq 4Seq 4
Seq 3Seq 3b1b1
b2b2
b5b5
b3b3
b4b4
virtual root: vrvirtual root: vr
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
![Page 30: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/30.jpg)
30
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
Seq 1Seq 1
Seq 2Seq 2 Seq 4Seq 4
Seq 3Seq 3b1b1
b2b2
b5b5
b3b3
b4b4
P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T) P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T)
mm
vrvr
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
![Page 31: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/31.jpg)
31
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
Seq 1Seq 1
Seq 2Seq 2 Seq 4Seq 4
Seq 3Seq 3b1b1
b2b2
b5b5
b3b3
b4b4
P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T) P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T)
mm
vrvr
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
![Page 32: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/32.jpg)
32
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
Seq 1Seq 1
Seq 2Seq 2 Seq 4Seq 4
Seq 3Seq 3b1b1
b2b2
b5b5
b3b3
b4b4
mm
vrvr
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T) P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T)
Floating-point & memory intensive
![Page 33: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/33.jpg)
33
Maximum Likelihood
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
Seq 1Seq 1
Seq 2Seq 2 Seq 4Seq 4
Seq 3Seq 3b1b1
b2b2
b5b5
b3b3
b4b4
mm
vrvr
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
Prior probabilities,Prior probabilities,Empirical base frequenciesEmpirical base frequencies
ππA A ππC C ππG G ππT T
P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T) P(A) P(C) P(G) P(T)P(A) P(C) P(G) P(T)
Sites evolve independently → we can compute per-site likelihoods in
parallel :-)
![Page 34: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/34.jpg)
Outline● Introduction to Phylogenetic Inference -
Alexandros● The RAxML Search Algorithm - Alexandros● Improvements in RAxML Next Generation -
Alexey● Tutorial - Alexey
![Page 35: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/35.jpg)
35
How does it work?
Compute randomized stepwise addition orderMaximum Parsimony tree
![Page 36: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/36.jpg)
36
How does it work?
Compute randomized stepwise addition orderMaximum Parsimony tree
Advantage of RAxML: search starts from distinct point in search space
every time
![Page 37: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/37.jpg)
37
How does it work?
Compute randomized stepwise addition orderMaximum Parsimony tree
Advantage of RAxML: search starts from distinct point in search space
every time
Alternatively, we can start froma completely random tree
![Page 38: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/38.jpg)
38
Starting Trees
good scores
bad scores
search space
Global maximum
Good starting tree
![Page 39: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/39.jpg)
39
Starting Trees
good scores
bad scores
search space
Global maximum
Another good starting tree
![Page 40: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/40.jpg)
40
How does it work?
Apply lazy subtree rearrangements
Compute randomized stepwise addition orderMaximum Parsimony tree
![Page 41: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/41.jpg)
41
How does it work?
Apply lazy subtree rearrangements
Compute randomized stepwise addition orderMaximum Parsimony tree
Iterate while tree improves
![Page 42: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/42.jpg)
42
Subtree Pruning & Re-Grafting
ST5
ST2
ST6ST4
ST3
ST1
![Page 43: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/43.jpg)
43
Subtree Pruning & Re-Grafting
ST5
ST2
ST6
ST4
ST3
ST1
prune
![Page 44: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/44.jpg)
44
ST5
ST2
ST6
ST4
ST3
ST1+1
Subtree Pruning & Re-Grafting
re-graft
![Page 45: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/45.jpg)
45
ST5
ST2
ST6
ST4
ST3
ST1+1
Subtree Pruning & Re-Grafting
![Page 46: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/46.jpg)
46
ST5
ST2ST6
ST4
ST3
ST1+1
Subtree Pruning & Re-Grafting
![Page 47: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/47.jpg)
47
ST5
ST2ST6
ST4
ST3
ST1+1
Subtree Pruning & Re-Grafting
![Page 48: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/48.jpg)
48
ST5
ST2
ST6 ST4
ST3
ST1+2
Subtree Pruning & Re-Grafting
![Page 49: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/49.jpg)
49
ST5
ST2
ST6 ST4
ST3
ST1+2
Subtree Pruning & Re-Grafting
![Page 50: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/50.jpg)
50
ST5
ST2
ST6 ST4
ST3
ST1+2
This search parameter can be explicitly set in RAxML-NG via
--spr-radius
Subtree Pruning & Re-Grafting
![Page 51: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/51.jpg)
51
ST5
ST2
ST6 ST4
ST3
ST1+2
If you don't set it RAxML-NG will auto-tune it
Subtree Pruning & Re-Grafting
![Page 52: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/52.jpg)
52
ST5
ST2
ST6 ST4
ST3
ST1
Optimize all branches!
Subtree Pruning & Re-Grafting
![Page 53: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/53.jpg)
53
ST5
ST2
ST6 ST4
ST3
ST1
Do we need to optimize all branches?
Subtree Pruning & Re-Grafting
![Page 54: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/54.jpg)
54
ST5
ST2
ST6
ST4
ST3
ST1
Lazy Subtree Pruning & Re-Grafting
![Page 55: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/55.jpg)
55
ST5
ST2
ST6
ST4
ST3
ST1
Lazy Subtree Pruning & Re-Grafting
“FAST” SPR round: no branches are optimized
![Page 56: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/56.jpg)
56
ST5
ST2
ST6
ST4
ST3
ST1
Lazy Subtree Pruning & Re-Grafting
“SLOW” SPR round: three adjacent branches are optimized
![Page 57: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/57.jpg)
Outline● Introduction to Phylogenetic Inference -
Alexandros● The RAxML Search Algorithm - Alexandros● Improvements in RAxML Next Generation -
Alexey● Tutorial - Alexey
![Page 58: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/58.jpg)
58
Evolution of RAxML(-NG)
2006 2013 2017
RAxML-NG
ExaML
RAxML
RAxML-Light
2019 202020152003
III VI v8.0
v0.1 v0.8
v1.0 v3.0
v8.2.12
v3.0.21
2021
v0.9 v1.0
2006 2013 2017
RAxML-NG
ExaML
RAxML
RAxML-Light
2019 202020152003
III VI v8.0
v0.1 v0.8
v1.0 v3.0
v8.2.12
v3.0.21
2021
v0.9 v1.0
![Page 59: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/59.jpg)
59
RAxML-NG design goals● Full rewrite of RAxML
– Search heuristic largely unchanged (as of v1.0)
● Improve maintainability & enable code reuse● Eliminate known bugs & bottlenecks● Improve user experience
– by default, “do the right thing”
![Page 60: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/60.jpg)
60
RAxML-NG & family
libpll / pll-modules
RAxML-NG
EPA-NG
ModelTest-NG
Vectorized PLF kernels(SSE3, AVX, AVX2)
Parsimony
MSA validation& statistics
Discrete tree operations
RAxMLFile I/O
(MSA & trees)
Numerical optimization(Stamatakis ’06, ‘14)
(Kozlov ’19)
(Darriba ‘20)
(Barbera ’18)
(Flouri in prep.)
GeneRax(Morel ‘20)
![Page 61: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/61.jpg)
61
Improvements & new features● Flexible evolutionary models
– All “classical” + custom DNA models (eg. DNA010010 = HKY)– Per-partition rate heterogeneity (incl. FreeRate)– Proportional branch lengths
● Phylogenetic terrace detection (Biczok ‘17)
● Transfer bootstrap support metric (Lemoine ‘18)
● Energy monitoring
![Page 62: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/62.jpg)
62
Performance & scalability● Checkpointing● Advanced load balancing● Binary alignment format● “Site repeats” optimization (Kobert ‘17)
– 10-60% speedup + memory savings
● Flexible and user-friendly parallelization
from ExaML
![Page 63: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/63.jpg)
63
Parallelization: hardware
CPU Vectorization (SSE, AVX ...)
Desktop Multi-threading
Cluster MPI
raxml-ngraxmlHPC
raxmlHPC-AVX
raxmlHPC-MPI
raxmlHPC-HYBRID
raxmlHPC-MPI-AVX2
raxmlHPC-SSE
raxmlHPC-PTHREADS
raxmlHPC-PTHREADS-SSE
raxml-ng-mpi
![Page 64: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/64.jpg)
64
Parallelization: software
Fine-grained Coarse-grained Mixed/hybrid
New in v1.0: Full native support and automatic configuration!
sites
T1
T2
T3
T4
T1
T2
T3
T4
T1
T3 T4
T2
T1 T2
T3 T4
T1 T2 T3 T4
T1 T2 T3 T4
T1 T2 T3 T4
T1 T2 T3 T4
Search 1
Search 2
Search 3
Search 4
Search 1
Alignment 4 threads 4 searches
T1, T2, T3, T4 e.g. from 4 starting trees
![Page 65: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/65.jpg)
Outline● Introduction to Phylogenetic Inference -
Alexandros● The RAxML Search Algorithm - Alexandros● Improvements in RAxML Next Generation -
Alexey● Tutorial - Alexey
![Page 66: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/66.jpg)
66
Quick start: ML tree search● Default command: --search
– 20 starting trees (10 random + 10 parsimony)– Pick the best-scoring one
$ raxml-ng --msa prim.phy --model GTR+G
Seq1Seq1Seq2Seq2Seq3Seq3Seq4Seq4
AlignmentAlignment
Length: mLength: m
AACCGGTT
A C G TA C G T
SubstitutionSubstitutionmodelmodel
![Page 67: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/67.jpg)
67
ML tree search: OutputAnalysis options: run mode: ML tree search start tree(s): random (10) + parsimony (10)...Starting ML tree search with 20 distinct starting trees
[00:00:00 -7871.515760] Initial branch length optimization...[00:00:00 -5736.644605] FAST spr round 1 (radius: 5)...[00:00:00 -5709.394601] SLOW spr round 1 (radius: 5)...[00:00:00] ML tree search #1, logLikelihood: -5708.979717...[00:00:07] ML tree search #20, logLikelihood: -5709.014076
Final LogLikelihood: -5708.923977
Best ML tree saved to: /home/alexey/test/prim.phy.raxml.bestTreeOptimized model saved to: /home/alexey/test/prim.phy.raxml.bestModel
--log info to hide search progress
![Page 68: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/68.jpg)
68
Tree with support values● All-in-one mode: --all
– ML tree search (as before)– Bootstrapping with autoMRE convergence test– Compute support values + map on ML tree
$ raxml-ng --all --msa prim.phy --model GTR+G --prefix A1 --seed 1
Output files: A1.raxml.* Fixed RNG seedfor better reproducibility
![Page 69: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/69.jpg)
69
Tree with support values: OutputAnalysis options: run mode: ML tree search + bootstrapping (Felsenstein Bootstrap) start tree(s): random (10) + parsimony (10) bootstrap replicates: max: 1000 + bootstopping (autoMRE, cutoff: 0.030000)
Starting ML tree search with 20 distinct starting trees...[00:00:02] ML tree search completed, best tree logLH: -5708.926130
[00:00:02] Starting bootstrapping analysis with 1000 replicates....[00:00:14] Bootstrapping converged after 100 replicates.
Best ML tree with Felsenstein bootstrap (FBP) support values saved to: /home/alexey/test/A1.raxml.support...Bootstrap trees saved to: /home/alexey/test/A1.raxml.bootstraps
![Page 70: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/70.jpg)
70
Customize analysis● Starting trees: --tree
● Bootstrap replicates: --bs-trees
● Branch length linkage mode: --brlen
● Branch length limits: --blmin / --blmax
--tree rand{1} --tree pars{50},rand{50} --tree user.nw
--bs-trees 100 --bs-trees autoMRE{500} --bs-trees bs.bw
--brlen linked --brlen scaled --brlen unlinked--brlen linked
--blmin 1e-9 --blmax 10
![Page 71: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/71.jpg)
71
Tree likelihood evaluation● Optimize free model parameters and branch
lengths on a fixed topology: --evaluate$ raxml-ng --evaluate --msa prim.phy --tree A1.raxml.bestTree --model GTR+G --prefix E1
Evaluating 1 trees
[00:00:00] Tree #1, initial LogLikelihood: -6419.996676
[00:00:00 -6419.996676] Initial branch length optimization[00:00:00 -6276.983451] Model parameter optimization (eps = 0.100000)
[00:00:00] Tree #1, final logLikelihood: -5709.005148...Best ML tree saved to: /home/alexey/test/E1.raxml.bestTreeOptimized model saved to: /home/alexey/test/E1.raxml.bestModel
![Page 72: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/72.jpg)
72
Tree likelihood evaluation (2)● Compute and print tree log-likelihood: --loglh
– No branch length optimization – No model optimization– No output files created
$ raxml-ng --loglh --msa prim.phy --tree A1.raxml.bestTree --model GTR{1/2/3/4/5/6}+G{0.5}
Final LogLikelihood: -6334.023267
![Page 73: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/73.jpg)
73
Check & parse● Check alignment for format errors: --check
● Compress alignment into binary file: --parse
● ...which can be then used in parallel jobs
$ raxml-ng --check --msa prim.phy –model GTR+G
$ raxml-ng --parse --msa prim.phy --model GTR+G --prefix prim
$ raxml-ng --search --msa prim.raxml.rba --prefix S1
![Page 74: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/74.jpg)
74
Parallelization tuning● Fully automatic (default) → heuristic-based$ raxml-ng --msa prim.rba
System: Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz, 16 cores, 62 GB RAM...Analysis options: run mode: ML tree search start tree(s): random (10) + parsimony (10)... parallelization: coarse-grained (auto), PTHREADS (auto)...[00:00:00] Alignment comprises 12 taxa, 1 partitions and 413 patterns...Parallelization scheme autoconfig: 16 worker(s) x 1 thread(s)...[00:00:00] Data distribution: max. partitions/sites/weight per thread: 1 / 413 / 6608[00:00:00] Data distribution: max. searches per worker: 2
![Page 75: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/75.jpg)
75
Parallelization tuning● Automatic with upper limits
● Manual
● Also works with MPI$ mpirun -n 4 raxml-ng-mpi --msa prim.rba --threads 16 -–workers 8
$ raxml-ng --msa prim.rba --threads auto{16} -–workers auto{2}
$ raxml-ng --msa prim.rba --threads 16 -–workers 2
4 ranks * 16 threads = 64 = 8 workers * 8 threads
![Page 76: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/76.jpg)
76
Energy monitoring● New in RAxML-NG v1.0: energy usage report
– Measured with Intel RAPL → CPU+DRAM only– Supported on Linux systems only – To disable, add: --extra energy-off
![Page 77: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/77.jpg)
77
Energy monitoring● New in RAxML-NG v1.0: energy usage report
– Measured with Intel RAPL → CPU+DRAM only– Supported on Linux systems only – To disable, add: --extra energy-off
Elapsed time: 42846.287 seconds
Consumed energy: 162370.469 Wh (= 812 km in an electric car, or 4059 km with an e-scooter!)
Single tree search (96 nodes x 12h): >160 kWhMy apartment per month: ~100 kWh
![Page 78: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/78.jpg)
78
You can't improve what you can't measure!
![Page 79: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/79.jpg)
79
You can't improve what you can't measure!
v1.0+ v0.9
-30%
default:
![Page 80: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/80.jpg)
80
RAxML-NG availability● Web services
– Vital-IT: https://raxml-ng.vital-it.ch/#/ → free, for small datasets– CIPRES: http://www.phylo.org/ → registration required
● Source+binaries for Linux & macOS– GitHub: https://github.com/amkozlov/raxml-ng
● Conda: https://anaconda.org/bioconda/raxml-ng● GUI: https://github.com/AntonelliLab/raxmlGUI
![Page 81: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/81.jpg)
81
Where to get help?● Documentation
https://github.com/amkozlov/raxml-ng/wiki
● Tutorialhttps://github.com/amkozlov/raxml-ng/wiki/Tutorial
● User support grouphttps://groups.google.com/forum/#!forum/raxml
![Page 82: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/82.jpg)
Q & A
![Page 83: Alexandros Stamatakis and Alexey Kozlov · 2021. 1. 13. · RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference Computational Molecular](https://reader036.vdocument.in/reader036/viewer/2022071302/60b15b304a4cda64445ada96/html5/thumbnails/83.jpg)
83
References● Barbera et al. (2018) EPA-ng: Massively Parallel Evolutionary Placement of Genetic Sequences. Systematic Biology,
syy054, https://doi.org/10.1093/sysbio/syy054● Biczok et al. (2017) Two C++ libraries for counting trees on a phylogenetic terrace. Bioinformatics.
https://doi.org/10.1093/bioinformatics/bty384● Kobert et al. (2017) Efficient detection of repeating sites to accelerate phylogenetic likelihood calculations. Syst. Biol
https://doi.org/10.1093/sysbio/syw075● Kozlov et al. (2019) RAxML-NG: A fast, scalable, and user-friendly tool for maximum likelihood phylogenetic
inference. Bioinformatics. https://doi.org/10.1093/bioinformatics/btz305● Kozlov, Aberer, Stamatakis (2015) ExaML version 3: a tool for phylogenomic analyses on supercomputers.
Bioinformatics. https://doi.org/10.1093/bioinformatics/btv184● Lemoine et al. (2018) Renewing Felsensteinen phylogenetic bootstrap in the era of big data. Nature.
https://doi.org/10.1038/s41586-018-0043-0 ● Morel, Kozlov, Stamatakis (2019) ParGenes: a tool for massively parallel model selection and phylogenetic tree
inference on thousands of genes. Bioinformatics. https://doi.org/10.1093/bioinformatics/bty839● Morel et al (2020). GeneRax: A tool for species tree-aware maximum likelihood based gene tree inference under gene
duplication, transfer, and loss. MBE. https://doi.org/10.1093/molbev/msaa141 ● Stamatakis (2006) RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with thousands of taxa and
mixed models. Bioinformatics, https://doi.org/10.1093/bioinformatics/btl446● Stamatakis (2014) RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies.
Bioinformatics, https://doi.org/10.1093/bioinformatics/btu033● Stamatakis and Aberer (2013) Novel parallelization schemes for large-scale likelihood-based phylogenetic inference.
In Parallel Distributed Processing (IPDPS) https://doi.org/10.1109/IPDPS.2013.70