Platform Approach to Cancer Discovery and Diagnostics Development

Transcription

1 Platform Approach to Cancer Discovery and Diagnostics Development Joanne Yeakley Staff Manager, Scientific Research Diagnostic Assay Development 2009 Illumina, Inc. All rights reserved. Illumina, illuminadx, Solexa, Making Sense Out of Life, Oligator, Sentrix, GoldenGate, GoldenGate Indexing, DASL, BeadArray, Array of Arrays, Infinium, BeadXpress, VeraCode, IntelliHyb, iselect, CSPro, and GenomeStudio are registered trademarks or trademarks of Illumina, Inc. All other brands and names contained herein are the property of their respective owners Illumina, Inc. All rights reserved. Illumina, illuminadx, Solexa, Making Sense Out of Life, Oligator, Sentrix, GoldenGate, GoldenGate Indexing, DASL, BeadArray, Array of Arrays, Infinium, BeadXpress, VeraCode, IntelliHyb, iselect, CSPro, GenomeStudio, Genetic Energy, and HiSeq are registered trademarks or trademarks of Illumina, Inc. All other brands and names contained herein are the property of their respective owners.

2 Sales ($ millions) Oncology is a Key Growth Segment Molecular Diagnostics $5, % Growth 16% Pharmacogenomics $200; 32% Oncology $600; 27% $2,375 Histo./Cytology $700; 23% Source: The World Wide Market for In Vitro Diagnostics Tests 5th Edition

3 The US Population Male United States: 2000 Female Population (millions) 3

4 is Aging Male United States: Female Population (millions) 4

5 Resulting in New Chronic Diseases Life Expectancy Top 10 Incurable diseases 1. Common Cold Sporadic 2. Cancer 3. Asthma 4. HIV 5. Diabetes 6. CJD - Lethal 7. Influenza Vaccines/Lethal 8. Lupus 9. Polio - Vaccines 10. Ebola - Lethal Cancer Asthma HIV Diabetes Lupus World Resources Institute

6 In Spite of Improvements Annual Age-adjusted Cancer Death Rates* Among Males for Selected Cancers, United States, 1930 to 2004 From Jemal, A. et al. CA Cancer J Clin 2008;58: Copyright 2008 American Cancer Society 6

7 Cancer Remains Lethal Heart disease 97.4% Cancer Stroke Chronic Respiratory Diabetes Few Established NATs Future technology fit Compelling unmet needs Causes of Death (USA, 2003) All other Diabetes Resp. Stroke Cancer Heart Accident Source: TIME, December 4, 2006, citing CDC and National Transportation Safety Board 7

8 Cancer Therapies Dominate the Biotech Pharma Pipeline Solid Tumors Lead Cancer Therapies Cancer 210 Solid Tumors 145 Infectious Disease AutoImmune Unspecified Cancers Lung Leukemia CVD 22 Prostate 79 HIV Other Breast ColoRectal Lymphoma Neurological Disease Diabetes Other Cancers Skin Pancreatic Digestive Disorders 14 Cancer-Related Conditions 39 Respiratory Disease Blood Disorders Kidney Brain Ovarian Genetic Disorders 9 Myeloma 31 Skin Disorders Eye Conditions 7 6 Head/Neck Liver Stomach Growth Disorders Transplantation 4 4 Cervical Bladder Sarcoma Source: PhRMA: 2006 Medicines in Development Biotechnology

9 Probability of Cure (%) Oncology Discovery Initial Targets Testis Thyroid 70 Melanoma Breast 60 Bladder Prostate 50 Kidney Oral Cervix/ Uteri Colon/ rectum 40 Ovary NHL Nasopharynx CNS Brain Myeloma Leukemia Esophagus Gastric Source: WHO 10 0 Lung Pancreas Liver Percentage of All Cancers 9

10 Oncology Discovery Ovarian Cancer Deadly 204,449 New cases annually; 124,860 deaths Incurable 110% 90% 92% Cum % of Cases 5 Yr Survival 85% 82% 100% Less than 40% are cured 70% 69% 72% Difficult Dx Patients present with a suspicious/palpable mass Early Dx is Key Five year survival is good if diagnosed early, but most patients are diagnosed late stage 50% 30% 10% 19% 22% 30% 56% 33% 37% 51% 40% 47% 39% 54% 26% 17% 12% Illumina Solution Develop a diagnostic assay which will diagnose ovarian cancer at an early stage -10% Ia Ib Ic Iia Iib Iic IIIa IIIb IIIc IV 10

11 Two Factors Make Ovarian Cancer Lethal Resistance to Therapy Due to Interference with the Apoptotic Pathway Mitochondria Cytochrome c Bcl-x(L) RESPONSE TO SALVAGE TREATMENT Step 1. Cytochrome c release Bad 60% Cytochrome c Apaf-1 Step 2. Apoptosome Formation Procaspase-9 AKT 33% 25% Procaspase-9 XIAP Step 3. Caspase Activation HtrA1 <6 6 to to 24 >24 Overexpression Down Regulation Functional apoptosome with active caspase-9 Time between last chemotherapy and relapse Source: Ovarian Cancer Coukos, G, et al. 11

12 12 A Disruptive Sequencing Technology

13 Cost per Human Genome The Tipping Point in Human Genome Sequencing $100M Costs Venter Genomes 1M $10M Watson 100k $1M African, Asian, Cancer pair 10k $100k $10k 169 in Genbank 1,000 Individual Genome Sequencing Time 13

14 SBS Sequencing Technology 3 5 DNA (<1 ug) Sample preparation Single Cluster molecule growtharray A C T C T G C T G A A G 5 T G C T A C G A T A C C C G A T C G A T Sequencing T G C T A C G A T Image acquisition Base calling 14

15 Platform Features Feature GAIIx HiSeq2000 Flowcells x Surface Imaging 1 x 1 2 x 2 Read length 2 x x 100 Yield per run (PF data) 50 Gb Gb Data Rate 5 Gb / day Gb / day Human Genomes (30x) per run 0.5 >2 15

16 Empirical Q score Average Q score Data Quality Q scores are accurate >90% bases are>99.9% accurate (2 x 100) Predicted Q score Cycle Average accuracy (aligned PF data) 99.95% 99.93% 99.90% 99.79% % PF bases > Q30 (99.9% accuracy) 96.5% 95.9% 94.6% 92.3% 16

17 Applications Whole genome re-sequencing Targeted sequencing (regions, genes, exomes) de novo sequencing Whole transcriptome sequencing mirna discovery and profiling DNA Methylation Histone Modification DNA-protein interaction Others 17

18 Whole Genome Sequencing: In House Projects Instrument Readlength (bp) Gb (PF) per run Flowcells per genome Project GA Yoruba genomes GA II genomes pilot GA II Melanoma, PGS pilot GA IIx Cancer, PGS GA IIx Cancer, PGS HiSeq <1 Human, Cancer 100 genomes completed to >30x to date in house using the GA 18

19 First African Individual Genome Sequence defines personal genetic profile Sequence data covers the 3 billion base genome to a depth of 40x 99.9% coverage at >99.999% consensus accuracy SNPs: 3.8 million (26% novel) Structural variants: 0.4 million Variation reflects ethnic origin and natural selection Near-complete individual profile of genetic variants Bentley et al. Nature 456: pp (2008) 19

20 Impact of Human Whole Genome Sequencing A Whole Genome Approach: Comprehensive catalogues of variation Discovery, characterisation and screening Address all types of sequence variant Germline and somatic mutation discovery Understand disease mechanisms Future applications in healthcare, personalised medicine 20

21 21 Cancer Whole Genome Sequencing Pilot

22 Whole Genome Sequencing: In House Projects Instrument Readlength (bp) Gb (PF) per run Flowcells per genome Project GA Yoruba genomes GA II genomes pilot GA II Melanoma, PGS pilot GA IIx Cancer, PGS GA IIx Cancer, PGS HiSeq <1 Human, Cancer 100 genomes completed to >30x to date in house using the GA 22

23 Melanoma: COLO-829 Collaboration with Mike Stratton, et Sanger Melanoma cell line COLO-829 and its matched normal (>30x depth) Male; metastatic tumour; prior to treatment BRAF, CDKN2A and PTEN mutations previously confirmed by exon sequencing SKY karyotype indicates pseudotriploidy Karyotype courtesy of Paul Edwards, University of Cambridge 23

24 Variant Detection Run specifications 2 x 75 base runs, 15 Gb / run. 0.2, 1, 2 & 4 kb libraries Single base substitutions Paired reads aligned with ELAND (non-gapped) Requires Q100 per allele, e.g. 3 x Q33 bases or equivalent Small indels Split reads aligned individually or grouped (Pindel, BWA, Grouper) Minimum 3x depth Larger structural variants Anomalous read pairs from short and long insert libraries >3 S.D. (deletions) or >5 S.D. (insertions) Minimum 10 reads Copy number changes Read depth (BWA+) in 10 kb windows 24

25 Somatic Substitutions in COLO-829 Subtraction process High stringency calibrated tumour calls (germline + somatic) Subtract if evidence of variant in normal; & dbsnp 33,345 somatic substitutions (gain of an allele) in total 1 per 100 kb 32,325 single base 510 double base (adjacent) 96.6% specificity 454/470 novel somatic calls confirmed by new PCR-capillary sequencing 16 may be artefacts of either method 88% (98%) sensitivity 42/48 previous somatic variants called automatically (98%) or 47/48 automatic + manual (five seen only manually as the evidence was below automatic threshold) 1/48 disputed call 25

26 Somatic Mutational Signature of UV Exposure 69.9% of all somatic substitutions are C T or G A 92% of somatic C T are in a YC pyrimidine dimer (vs. 53% by chance) T G C A 26

27 Complex CNV in tumour on chromosome 3 Tumor Tumour (40X) Normal (30X) Normal 4 genes in this region: RARB, TOP2B, NGL1,KS 27

28 Catalogue of COLO-829 Somatic Variants 33,345 Single base substitutions 286 coding 1018 small indels 14 coding 37 Structural rearrangements 34 intrachromosomal: 25 deletions 6 insertions 2 duplications 1 complex 3 interchromosomal 19 breakpoints in genes 198 changes in copy number Ideograms / val indels / sn 10M / cdsn / cn /loh / sv 28

29 29 Cancer Discovery 2010 Update

30 Whole Genome Sequencing: In House Projects Instrument Readlength (bp) Gb (PF) per run Flowcells per genome Project GA Yoruba genomes GA II genomes pilot GA II Melanoma, PGS pilot GA IIx Cancer, PGS GA IIx Cancer, PGS HiSeq <1 Human, Cancer 100 genomes completed to >30x to date in house using the GA 30

31 Ovarian Cancer Discovery Establish end-to-end workflow for sequencing, analysis, interpretation Refine analysis tools Sequence 25 tumour normal genome pairs (>30x depth) Identify genome wide catalogues of somatic mutation Search for mutational signatures, novel biomarkers, etc. Incorporate RNA and methylation data Gain experience in line with ICGC and similar projects 31

32 Illumina s Oncology Discovery Initiative Samples Discovery Validation Dx Service Dx Product Clinical Partnership Obtain consent Obtain samples Prep samples Build libraries 25 Tumor/ Normals W GS Methylome Transcriptome samples BeadChip analysis LDT in CLIA lab Prospective trials CE- IVD US IVD Ovarian is here 32

33 Cancer Sequencing Workflow Receive samples Genotype for tracking and QC Sequence & QC data Build genome (CASAVA) Check tracking and QC Call somatic variants: - substitutions - indels <100 bp - SVs >100 bp - CNVs Annotate variants Cross-sample comparison Visualize in IGV 33

34 Variant Detection Run specifications 300 bp lib, 2x100s, PF Gb/fc, revised (v4) chemistry Revised (PL 1.6) image analysis, basecall calibration (>90% Q30 bases) Single base substitutions Paired reads aligned with ELAND2 (multi-seed, gapped alignment) Threshold 10, i.e. requires 3 x Q33 bases or equivalent Small indels Local assembly, gapped alignment of consensus (Grouper) Minimum 5x depth. Now seeing up to 100-base events Larger structural variants Anomalous read pairs from short and long insert libraries >3 S.D. (deletions) or >5 S.D. (insertions) Minimum 5 reads then additional evidence-based ranking Copy number changes Multiple approaches incl. recursive partitioning of read depth 35

35 Cancer Genome Summary Statistics Tumour Tumour depth Normal depth Somatic gains Melanoma x 32.5x 33,345 Breast x 29.0x 4,838 Ovarian A 40.5x 37.3x 7,147 Ovarian B 32.5x 35.1x 1,099,481 Ovarian C 38.0x 39.1x 4,284 Ovarian D 32.7x 37.2x 1,244 Ovarian E 36.2x 35.4x 3,727 Ovarian F 37.2x 29.7x 11,984 Ovarian G 30.4x 32.0x 3,091 Ovarian H 39.9x 35.0x 5,623 Ovarian I 34.8x 34.2x 8,824 Ovarian J 35.8x 39.9x 4,815 Tumour Tumour depth Normal depth Somatic gains Ovarian K 32.3x 33.0x 3,255 Ovarian L 32.8x 39.6x 3,093 Ovarian M 37.6x 42.9x 4,480 Ovarian N 38.3x 37.8x 6,698 Ovarian O 36.7x 30.1x 5,485 Ovarian P 35.9x 39.2x 3,489 Ovarian Q 34.7x 35.5x 5,584 Ovarian R 35.7x 33.9x 5,973 Ovarian S 34.5x 33.6x 2,154 Ovarian T 35.2x 33.3x 2,203 Ovarian U 38.1x 35.6x 2,881 36

36 IGV visualisation of two variants in TP53 Tumour L Normal L Tumour R Normal R 37

37 Summary of TP53 somatic substitutions (SNVs) 14/17 high grade serous samples have TP53 somatic substitution variants (SNVs) further analysis of remaining samples ongoing Review vs. COSMIC: All are in known TP53 hotspots All are novel observations in ovarian cancer Two seen previously in other cancers TP53 mutations reported in 97% of high grade pelvic serous carcinoma samples in a recent study (Ahmed et al 2010) 38

38 Copy Number: Chr6 C6orf170, FABP7, GJA1, HSF2, NKAIN2,PKIB, RLBP1L2, SERINC1, SLC25A5P7, SMPDL3A, TRDN sample100 sample102 Tumor Normal AIM,1 ASC3C, ATG5, BEND3, BVES, C6orf112, C6orf203, CCNC, GRIK2, HACE1, LIN28B, MCHR2, MIRN587, POPDC3, PRDM1, PRDM13, PREP, QRSL1, RPL35P3, RTN4IP1, SIM1, USP45 sample104 sample101 sample103 sample111 39

39 Circos 5 Tumor/Normal Pairs Genomes are extensively rearranged, even at early stages Tumors are heterogeneous Lots of fertile ground for biomarker discovery 40

40 Summary and Future Prospects A Genome-Wide Opportunity: Comprehensive variant catalogues >95% specificity and ~90% sensitivity of somatic substitutions System developments enable large-scale studies: v5 chemistry, software improvements (>99.9% raw data accuracy) HiSeq2000 platform (>300 Gb per run: two 30x genomes per instrument per week) Implications for individual genome sequencing: Understanding disease mechanism, biomarker discovery Sequencing individual cancer genomes for personalised treatment 41

41 Many Thanks to all our Collaborators and the Illumina teams... 42