clinical-grade next-generation sequencing: … next-generation sequencing: assay design &...
Post on 29-Nov-2018
218 Views
Preview:
TRANSCRIPT
Clinical-Grade Next-Generation Sequencing:
Assay Design & Validation
Richard D. Press, MD, PhDDept of Pathology
Knight Cancer InstituteKnight Diagnostic Labs
Oregon Health & Science UniversityPortland, OR
NGS Assay Design ConsiderationsCustomization is Key
• Coverage: medium, large, or X-large
• If not whole exome / genome, then:– Which genes to sequence? – Which regions?
• What cancers to cover?– Organ-specific?– Carcinoma vs sarcoma vs
hematopoeitic?– All cancers?
What size are you?
Assay Validation Questions(I don’t have the answers)
• How to validate all known mutation possibilities?
• Tumor cellularity requirements?• Minimal mutant allele burden detection?• Quantitative or Qualitative reporting?• Quality control: read depth? Other
parameters?• Unknown variants?• Many other questions and variables
What’s All the Fuss About NGS?- Broader coverage, including tumor suppressors- Better sensitivity, through deeper reads- ? Lower costs, through multiplexing- More comprehensive cancer genome characterization for
targeted therapeutic (and diagnostic) discovery
Single geneassays
ABI
Multiplexedhotspots
Multigenepanels
Ion Torrent PGMSequenom MassArray
Wholeexome
454
Wholegenome
Illumina HiSeq
Targeted Therapeutics
SHC GRB2SOS
NRASKRAS
BRAF
MEK
ERKPSTAT
P
PP
P
P PP
PP
P
PI3KPDK
AKT
P
mTOR
S6K
P
P
P
PTEN
Receptor tyrosine kinases
p53
ALK
ErlotinibLapatinibPF299804AfatinibImatinib RAF265
Vemurafenib
AZD6244PD0325901ARRY162
EverolimusTemsirolimus
BEZ235
BKM120BGT226BYL719
Crizotinib
MK2206SR13668
Coverage: How High Do You Want to Go?
• Mutation Hotspots• Disease-Specific
Gene Panels• Generic “Cancer”
Gene Panels• Whole Exome• Whole Genome
Higher Coverage means:• More complexity & cost• More unknown variants• ? Overkill for clinical
care: who cares if its not drugable?
Ion Torrent PGM
• Moderate throughput• Massively parallel 3rd
generation sequencing• Performed on a semi-
conductor chip
Ion Torrent PGM
Each well is a tiny solid-state pH meter
Library Preparation:Hybridization-Capture Approaches
NimblegenAgilentIon Torrent
50 to 1,000 ngGenomic DNA
Shear, End Repair,Ligate adapters
Gel purify at ~180 bp
Add biotinylated probesto genes/exons of interest
b
b
b
Purify hybridizedRNA probes with magnetic beads
b
b
Treat with NaOHto remove RNA Emulsion PCR
Amplify gDNA targets190-Amplicon multiplex PCRIon AmpliSeqTM Cancer Primer Pool
Remove genomic DNA template and primersPartially digest primers
Ligate barcode adaptersNick-translate and amplify
Genomic DNA
A BC
P1
A BC P1
Emulsion PCREnrichment of templated IonSphere particlesSequence
Ampliseq primer pool
Emulsion PCR
NGS Library Prep
Amplicon-Based Approach
AmpliSeq Cancer Panel• 739 hotspots covered by 190 amplicons• Single tube amplification • Average amplicon length: 119 bp (100-169 bp)• Input DNA: 10 ng (Fresh or FFPE)• Turn-around time: 48 hours • 46 genes:
ABL1, AKT1, ALK, APC, ATM, BRAF, CDH1, CDKN2A, CSF1R, CTNNB1, EGFR, ERBB2, ERBB4, FBXW7, FGFR1, FGFR2, FGFR3, FLT3, GNAS, HNF1A, HRAS, IDH1, JAK2, JAK3, KDR, KIT, KRAS, MET, MLH1, MPL, NOTCH1, NPM1, NRAS, PDGFRA, PIK3CA, PTEN, PTPN11, RB1, RET, SMAD4, SMARCB1, SMO, SRC, STK11, TP53, VHL
Mass Spectrometry-Based Detection of Genomic Mutations
PCR targets of interest~ 100 bp amplicons
Clean-up steps
Primer extension reaction
De-salting step
MALDI-TOF Mass spectrometry
AT
GC
TddA
CddG
G A
Mass spec profile for a 9-plex reaction
Multiplex of 9 assays
PIK3CA E542K wt 542Emutant 542Kunextendedprimer
Mass Spec Leukemia Panel: 370 mutations / 31 genes
ABL FLT3 KRASAKT1 FMS METAKT2 GATA1 MPLAKT3 HRAS NOTCH1BRAF IDH1 NPM1CBL IDH2 NRASCBLB JAK1 NTRK1FBXW7 JAK2 PAX5FES JAK3 PDGFRBFGFR4 KIT PTPN11
SOS1
Mutation Spectrum in AML (108 OHSU cases)
Normal cytogenetics: 78% mutation frequencyAbnormal cytogenetics: 43% mutation frequency
Mutation discovery with 31-gene mass spec panel plus single gene in-del assays (FLT3, CEBPA, KIT)
J Dunlap, in press
Ampliseq Validation Study
• 45 FFPE tumor DNA samples with known mutations previously quantitated on Sequenom MassArray (mass spectrometry)– 53 point mutations– 19 in/dels (range 4 - 63 bp)
• 7 unmatched FFPE normal tissue DNA• 100 bp single-end sequencing runs
– 22 samples run singly on 314 chips– All samples run as 4-plexes on 316 chips
TP53 R273H
PG1-48
Colonadenoc.
• General background is low• Some homopolymers (e.g. a run of C’s) lead to false positive calls
Read Coverage Distribution:190 Amplicons in each of 45 Tumor Samples
(Normalized to 400,000 reads)
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0
1
2
3
4
5
6
7
8
9
10
0-99
100-
199
200-
299
300-
399
400-
499
500-
599
600-
699
700-
799
800-
899
900-
999
1000
-109
911
00-1
199
1200
-129
913
00-1
399
1400
-149
915
00-1
599
1600
-169
917
00-1
799
1800
-189
919
00-1
999
2000
-209
921
00-2
199
2200
-229
923
00-2
399
2400
-249
925
00-2
599
2600
-269
927
00-2
799
2800
-289
929
00-2
999
3000
-309
931
00-3
199
3200
-329
933
00-3
399
3400
-349
935
00-3
599
3600
-369
937
00-3
799
3800
-389
939
00-3
999
4000
-409
941
00-4
199
4200
-429
943
00-4
399
4400
-449
945
00-4
599
4600
-469
947
00-4
799
4800
-489
9
Perc
enta
ge o
f tar
gets
≥ re
ad d
epth
Num
ber o
f tar
get r
egio
ns w
ith re
ad d
epth
Read depth
Average = 1948 reads
•95% of amplicons average >400 reads •91% of amplicons >650 reads (estimated 5% sensitivity) • 78% of amplicons with >1200 reads (estimated 1% sensitivity) Beadling, submitted 2012
Sequencing Performance
• 4-plex samples run on 316 chips (6.2 million wells)– Avg 4 million beads loaded– Avg 3.7 million beads had library templates (92%)– Avg 1.7 million beads yielded quality sequence
• For individual samples– Avg 428,000 reads/sample (range 178K - 710K)– Mean read length 76 bp– On-target reads: >95%
• Across all samples (normalized to 400K reads)– Avg 1,941 reads per amplicon – 95% of amplicons with > 400 reads
What is the Optimal Mutant Allele Burden Cutoff?53 known point mutations in 45 tumor samples
All 53 “detected” by both NGS & Mass Spectrometry
0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
0.80
0.90
1.00
0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00
True
pos
itive
rate
(sen
sitiv
ity)
False positive rate (1-specificity)
8% 1% variant allele5%
10%15%
50%
Receiver Operator Characteristic (ROC) Analysis
If lower limit of detection set at 8% mutant allele:- 100% Sensitivity- 94% Specificity
But are these few discordants really “false positives”?Or variants missed by less sensitive methods?
Beadling, submitted 2012
Summary of Variants• All 53 known point mutations were identified by the
variant caller software• 26 new mutations were also identified
– APC in colon ca; PTEN in endometrial ca; STK11 in lung ca
• 19 in/dels were included in the analysis; range: 4-63 bp – 2 called exactly– 5 flagged as point mutations– 12 visible on manual inspection but not flagged
• 54 ‘variants’ turned up reproducibly in both tumor and normal DNA, likely reflecting sequencing aberrations
Is NGS Data Quantitative?Allele Ratios: MassArray vs AmpliSeq
0%
20%
40%
60%
80%
100%
0% 20% 40% 60% 80% 100%
Red = 1:1 equivalence line
N = 53 mutationsSlope = 0.782Bias = minimal (3.4%) (Ampliseq lower than Sequenom)R = 0.92
Ion Torrent Ampliseq Allele Burden
Sequ
enom
Alle
le B
urde
n
(P < 0.0001)
Beadling, submitted 2012
NGS Clinical Validation: Status Report (OHSU; May ‘12)Ampliseq 46 gene panel (multiplex PCR library
prep)– Analytical validation essentially complete– 100% sensitive compared to “gold standards”– Variants of unknown function? – Low-level variants?
• We have validated an 8% mutant allele threshold– Indels remain a challenge for variant caller
• New software any better?– Aiming for summer ‘12 launch in our clinical lab
The Future: Disease-Specific Gene Panels
Cancer Site
Target genes
# Exons Kilobases Ampli-cons
New Genes(not in Ampliseq)
Lung 23 224 34.3 502 9
Colon 16 157 31.8 405 6
Melanoma 21 113 13.9 231 9
AML / ALL / MDS
42 342 62.6 863 28
• Ion Torrent custom primer design software used to design primers for library prep
• Proprietary primer modifications allow massive multiplexing: 2-4 PCR reactions per library prep
Many Unanswered Questions• How to re-validate assays given continuing
rapid pace of improvements to chemistries, hardware, and software?
• Sanger sequencing confirmation?• Unknown variants? • Matched normal tissue required?• Quality control? • Gene patent implications?• FDA? Can you image the approval process?• Will anyone pay us for this service?
AcknowledgementsOHSU Knight Diagnostic Labs
Carol Beadling, PhDTanaya NeffAndrea WarrickFei YangJennifer Dunlap MDChris Corless MD, PhD
Oregon Health & Science University
top related