Investigating the evolution of structural variation in cancer

PhD Completion Seminar.

Transcript

1. Completion Seminar Investigating the evolution of structural variation in cancer Supervisors: Prof. Chris Hovens, Dr. Niall Corcoran, Dr. Geoff Macintyre, Dr. Cheng-Soon Ong and A/Prof. Tony Papenfuss Marek Cmero, PhD Candidate - Dept. Surgery/CiS
2. Cancer is an evolutionary process
3. Thesis summary • Developed a method to study clonal evolution of genome rearrangements • Applied method to 2700+ whole-genome pan-cancer samples • Developed strategies to test for evidence of clonal change in low-quality longitudinal samples • Applied methods to study treatment resistance in aggressive prostate cancer cohort
4. SVclone A method for inferring structural variant cancer cell fraction
5. Different cancers are driven by different recurrent mutations Source: Ciriello et al. 2013, doi:10.1038/ng.2762
6. Using whole-genome sequencing data to elucidate intratumour heterogeneity (ITH)
7. chrA chrB Ref Donor Deletion Ref Donor Tandem duplication Ref Donor Translocation Ref Donor Interspersed duplication Ref Donor Inversion Inter-chromosomal translocation What is structural variation (SV)? Copy-number variation Balanced rearrangements chrA chrB Ref Donor Deletion Ref Donor Tandem duplication Ref Donor Translocation Ref Donor Interspersed duplication Ref Donor Inversion Inter-chromosomal translocation
8. What’s a VAF? Estimating population frequency from Variant Allele Frequency (VAF) TCTTCATTTGCAGCTAGTCGTGTGGGCTTAGT TCTTCATTTGCAGCTAGTCGTGTGGGCTTAGT TCTTCATTTGCAGCTAGTCGTGTGGGCTTAGT TCTTCATTTGCAGCTAGTCGTGTGGGCTTAGT TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT Coverage = 10 A: 6 REF G: 4 SNV VAF = 4/10 = 40% REF: TCTTCATTTGCAGCTAATCGTGTGGGCTTAGT
9. Calculating VAFs for SVs is complicated
10. Calculating VAFs for SVs is complicated Discordant spanning reads Split reads Normal reads
11. SV read counting
12. Solving read-count biases
13. Solving read-count biases
14. SVclone pipeline
15. What is cancer cell fraction? Normal cells Tumour (clone B) Tumour (clone A)
16. What is cancer cell fraction? Tumour (clone B) Tumour (clone A) CCF: 2/6 = 33.33’% CCF: 4/6 = 66.66’% We must infer clone CCFs from groups of variant CCFs
17. Copy-number confounds CCF calculation VAF = 1 / (2 + 3 + 4) = 11.11’% but CCF = 4 / 6 = 66.66’%
18. VAF to CCF Normal Tumour (variant) Tumour (reference)
19. Clustering using Bayesian Inference and MCMC • Define distributions for unknown parameters • Dirichlet process: number of clusters is dynamic • Iterative approach to simulate distributions that match the data using markov-chain monte- carlo (MCMC) • Extract CCFs and cluster memberships ϕj = population frequencies Zj = cluster gj = most likely copy-numbers dj = depth at variant locus bj = variant reads
20. Inferring ITH through CCFs
21. Visualization
22. Validating the clustering approach
23. Validating the clustering approach
24. SVclone addresses the gap for comprehensive DNA- based clonal deconvolution
25. Summary • Developed pipeline for inferring cancer cell fraction from structural variant breakpoint data, using: • Rule-based read counting • Clustering using Bayesian Inference • Validated using • Simulations • In-silico real sample mixtures • Only method available that: • considers balanced rearrangements • provides comprehensive clonal deconvolution of breakpoints
26. Application We apply the SVclone methodology to: • 2700+ pan-cancer samples • Aggressive prostate cancer cohort to study drug resistance
27. Pan-Cancer Analysis of Whole Genomes (PCAWG) https://dcc.icgc.org/pcawg
28. PCAWG SVclone results
29. PCAWG SVclone results
30. Intratumour heterogeneity has important clinical implications Source: Aparicio, S., & Caldas, C. (2013). The Implications of Clonal Genome Evolution for Cancer Medicine. https://doi.org/10.1056/NEJMra1204892
31. Case study: Neo-adjuvant prostate cancer cohort • 14 high-risk prostate cancer patients underwent neo- adjuvant ‘supercastration’ treatment • Biopsies were obtained prior to treatment and from a post- treatment prostatectomy • All pre-treatment samples were formalin fixed paraffin embedded (FFPE) • 27 samples underwent whole-genome sequencing • Coverage mean = 37.8x • Any evidence of clonal change in pre vs. post?
32. Neo-adjuvant cohort purity estimates Purity mean = 49.2%
33. No evidence of FFPE artefact from SNV calling
34. FFPE and fresh frozen samples have low VAF correlation
35. Simulating high-dispersion samples
36. Testing for subclonal shifts X bin Y bin count 0.00 – 0.25 0.5 – 0.75 56 0.00 – 0.25 0.75 – 1.00 83 0.00 – 0.25 1.00 – 1.25 44 0.25 – 0.50 0.75 – 1.00 44 0.25 – 0.50 1.00 – 1.25 28 0.50 – 0.75 1.00 – 1.25 37 X bin Y bin count 0.00 – 0.25 0.00 – 0.25 0 0.00 – 0.25 0.00 – 0.25 0 0.00 – 0.25 0.00 – 0.25 0 0.25 – 0.50 0.25 – 0.50 0 0.25 – 0.50 0.25 – 0.50 1 0.50 – 0.75 0.50 – 0.75 17 Pre < Post counts Pre > Post counts Apply 1-tailed unpaired T-test in direction of shift – results imply selection
37. Simulating clonal shifts
38. Simulation results Simulated shift beta (dispersion) Coverage Clonal selection FDR Clonal shrinkage FDR 0.1 > 0.9 1 35 **0.040 1.0000 0.2 > 0.8 1 35 0.2203 1.0000 0.3 > 0.7 1 35 0.1656 1.0000 0.4 > 0.6 1 35 0.3563 1.0000 0.5 > 0.5 1 35 0.5000 1.0000 0.1 > 0.9 10 35 **0.011 1.0000 0.2 > 0.8 10 35 **0.0001 1.0000 0.3 > 0.7 10 35 **0.0006 1.0000 0.4 > 0.6 10 35 **0.0155 1.0000 0.5 > 0.5 10 35 0.5000 1.0000 0.1 > 0.9 100 35 **0.0223 1.0000 0.2 > 0.8 100 35 **0.0044 1.0000 0.3 > 0.7 100 35 **0.011 1.0000 0.4 > 0.6 100 35 **0.0401 1.0000 0.5 > 0.5 100 35 0.3627 1.0000 0.1 > 0.9 1000 35 **0.0401 1.0000 0.2 > 0.8 1000 35 **0.0111 1.0000 0.3 > 0.7 1000 35 **0.0124 1.0000 0.4 > 0.6 1000 35 **0.0216 1.0000 0.5 > 0.5 1000 35 0.3813 1.0000
39. SCNA: FFPE vs. fresh frozen
40. Filtering SCNAs
41. SVclone – clustering FFPE vs. fresh frozen ** ** ** Poor purity estimates in RARP Good purity estimates
42. Pre vs. post SNV clustering results Poor purity estimates Selection?
43. SV filtering in FFPE samples Default filtering Require >1 spanning reads for inter-chromosomal translocations
44. SVclone FFPE vs. fresh frozen results ** Poor purity estimate
45. SVclone pre vs. post clustering results ** ** ** Poor purity estimates in RARP Selection?
46. SVclone pre vs. post clustering results
47. Summary • SVclone Pan cancer Analysis • Distinct SV/SNV heterogeneity patterns in diff. cancer types • Novel SCNR phenotype • Developed strategies to test for evidence of clonal change in low-quality longitudinal samples • Neo-adjuvant treatment cohort showed little evidence of changing clonal dynamics
48. Acknowledgements The University of Melbourne/Royal Melbourne Hospital • *Dr. Christopher Hovens • *Dr. Niall Corcoran • Natalie Kurganovs • Stefano Mangiola • Kangbo Mo Cancer Research UK Cambridge Institute • *Dr. Geoffrey Macintyre • *Dr. Florian Markowetz (host supervisor) Data61/Australian National University • *Dr. Cheng-Soon Ong The Walter + Eliza Hall Institute • *Dr. Tony Papenfuss • Dr. Jan Schröder The University of Glasgow • Dr. Ke Yuan PCAWG Evolution and Heterogeneity Working Group (esp. Peter Van Loo, David Wedge and Stefan Dentro) People (*supervisors) Funding Thank you to Roger Riordan of the Cybec Foundation for his generous Urology Fellowship funding contribution Compute Advisory panel • Dr. HongJian Zhu, Dr. Andrew Lonie SVclone: https://github.com/mcmero/SVclone Preprint on bioaRxiv coming soon!
49. Publications • 2017 • [Manuscript in preparation] Cmero, M., Ong, C.S., Yuan, K., Schröder, J., Kangbo, M., PCAWG Evolution and Heterogeneity Working Group, Corcoran, N.M., Papenfuss, T., Hovens, C.M., Markowetz, F., Macintyre, G. SVclone: inferring structural variant cancer cell fraction. • Gerstung, M. et al. [author 22 of 45] (2017). The evolutionary history of 2,658 cancers. bioRxiv 161562; doi: https://doi.org/10.1101/161562 • 2016 • Mangiola, S., Hong, M. K., Cmero, M., Kurganovs, N., Ryan, A., Costello, A. J., ... & Hovens, C. M. (2016). Comparing nodal versus bony metastatic spread using tumour phylogenies. Scientific Reports, 6. • Stuchbery, R., Macintyre, G., Cmero, M., Harewood, L. M., Peters, J. S., Costello, A. J., ... & Corcoran, N. M. (2016). Reduction in expression of the benign AR transcriptome is a hallmark of localised prostate cancer progression. Oncotarget, 7(21), 31384. • Sapre, N., Macintyre, G., Clarkson, M., Naeem, H., Cmero, M., Kowalczyk, A., ... & Hovens, C. M. (2016). A urinary microRNA signature can predict the presence of bladder urothelial carcinoma in patients undergoing surveillance. British journal of cancer, 114(4), 454-462. • 2015 • Johanson, T. M., Keown, A. A., Cmero, M., Yeo, J. H., Kumar, A., Lew, A. M., ... & Chong, M. M. (2015). Drosha controls dendritic cell development by cleaving messenger RNAs encoding inhibitors of myelopoiesis. Nature immunology, 16(11), 1134-1141. • Hong, Matthew K.H., et al. [author 9 of 29] (2015). Tracking the origins and drivers of subclonal metastatic expansion in prostate cancer." Nature communications 6 • 2014 • Johanson, T. M., Cmero, M., Wettenhall, J., Lew, A. M., Zhan, Y., & Chong, M. M. (2014). A microRNA expression atlas of mouse dendritic cell development. Immunology and Cell Biology. • Schroeder, J., Hsu, A., Boyle, S. E., Macintyre, G., Cmero, M., Tothill, R. W. & Papenfuss, A. T. (2014). Socrates: identification of genomic rearrangements in tumour genomes by re-aligning soft clipped reads. Bioinformatics, btt767.