Mutual exclusivity: drivers, pathways, and beyond Teresa Przytycka NIH / NLM / NCBI Cancer drivers, passengers, supporting actors, witnesses • Driver mutations /alterations– mutations contributing to cancer progression • Passenger mutations – neutral mutations accumulating during cancer progression • Challenges in detecting driver mutations: – Heterogeneity - phenotypically similar cancer cases might be caused by different sets of driver mutations – Rare drivers • Best supporting actors (Igor’s talk) • Witnesses (this talk) Cancer driving pathways examples of BRCA mutated genes in their pathway context RAS 1.5% or more PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 Mutual exclusivity of cancer drivers Thomas et al 2007 Ciriello, et al., 2012; Vandin, et al., 2012; Szczuret et.al , 2014, 2015 Leiserson, et al., Vadin et al. 2013,2014,2015; Kim et al. 2015 Constantinescu et al. 2015 patients mutations in gene 1 Mutations in gene 2 Explanations • Two drivers dysregulating the same pathway • Each of the drivers corresponds to of a unique cancer type or subtype • Negative genetic interactions between drivers 4 Mutually exclusive pairs often share pathways RAS PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 5 Fabio Vandin et al. Genome Res. 2012;22:375-385 Mutually exclusive pairs often share pathways RAS PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 6 Mutually exclusive pairs often share pathways RAS PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 7 Fabio Vandin et al. Genome Res. 2012;22:375-385 Mutually exclusive pairs often share pathways RAS PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 8 Fabio Vandin et al. Genome Res. 2012;22:375-385 Mutually exclusive pairs often share pathways RAS PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 9 Fabio Vandin et al. Genome Res. 2012;22:375-385 Mutual exclusivity relation with a gebe “outside” the pathway Genes exclusive with TP53 RAS PIK3CR1 PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 10 Fabio Vandin et al. Genome Res. 2012;22:375-385 Kim et al. 2016 Introducing a classification of mutual exclusivity Motivation – distinguishing ME between drivers that: • Result in a similar phenotype (WITHIN_ME) • Occur across multiple cancer types (ACROSS_ME) • Between type specific drivers (BETWEEN_ME) Kim et al. ISMB /Bioinformatics 2015 Mutual exclusivity classes in PanCancer context • Within tissue exclusivity WITHIN_ME Traditional permutation test • Across tissues exclusivity ACROSS_ME Type-restricted permutation test • Between tissues exclusivity BETWEEN_ME 12 Traditional, type-oblivious permutation test Kim et al. ISMB /Bioinformatics 2015 WITHIN and ACRPSS ME is biased towards pathway edges 13 Kim et al. ISMB /Bioinformatics 2015 Finding cross-cancer dysregulated modules by combining interaction and ACROSS_ME • Within tissue exclusivity WITHIN_ME Traditional permutation test • Across tissues exclusivity ACROSS_ME Type-restricted permutation test • Between tissues exclusivity BETWEEN_ME Type-oblivious permutation test 14 Finding PanCancer dysregulated pathway - ME Module Cover Approach Optimization problem: Find smallest cost set of modules so that each disease case is covered at least k times Cost is a function of: distance in the network of genes in same module Mutual exclusivity Score of covering edge Optimization problem: unit cost per module Gene cover Kim et al. PSB 2013, Bioinformatics 2015 Does putting together ACROSS_ME and interaction data actually helps ? MEMCover we find more cancer drivers Compared to Module Cover Compared to HotNet2 16 Robust mutual exclusivity within some modules 17 Kim et al. ISMB /Bioinformatics 2015 Hub-like ME within some modules 18 Splicing 19 Kim et al. ISMB /Bioinformatics 2015 No ME within some modules 20 Kim et al. ISMB /Bioinformatics 2015 Mutual Exclusivity Hubs 21 Kim et al. ISMB /Bioinformatics 2015 Beyond cancer drivers A. B. BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted TTN – presumed passenger - no known role in cancer A. B. BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted Presumed to be passenger mutations gene has a role in cancer A. B. BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted FBXW7 – tumor suppressor but can harbor passenger mutations A. B. FBXW7 BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted If TTN is a passenger that what is the train it is ridding on? A. B. BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted TTN carries APOBEC signature in BRCA and Pol ε signature in UCEC From Alexandrov et al, Nature 2013 Consistent with TTN spectrum in BRCA APOBEC cytidine deaminase mutational spectrum Consistent with TTN spectrum in UCEC Pol II ε mutation mutational spectrum 27 TTN and TP53 have common neighbors in BRACE A. B. BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted Co-occurrences - a causal relation or same underlying cause? A. B. FBXW7 BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted Can APOBEC cause TP53 mutations? Burns et al. TP53, TTN concurrence (p-value < 0.0002, hypergeometric test). 30 Can APOBEC cause TP53 mutations? Burns et al. TP53, TTN concurrence (p-value < 0.0002, hypergeometric test). TP53, TTN concurrence after correcting for patients mutation frequency p-value > 0.29 31 Can APOBEC cause TP53 mutations? Burns et al. TP53, TTN concurrence Immune response (p-value < 0.0002, hypergeometric test). TP53, TTN concurrence after correcting for APOBEC patients mutation frequency p-value > 0.29 TP53 TTN 32 True for all TP53 mutations in BRCA? NO Immune response APOBEC TP53 TTN 33 Co-occurrences - a causal relation or same underlying cause? A. B. FBXW7 BRCA (FDR 0.0125) UCEC (FDR 0.0025) (computed with our new method WeSME; width p-value; color shade FDR) Kim et al. submitted TTN and TP53 share exclusivity partners TP53 RAS TTN PTEN PIK3CA CTCF AKT1 MAPK signaling FOXA1 MAP3K1 MAP3K4 SWI/SNF MAP2K4 ARID1A mTOR NCOA3 EP300 CEBPA Mediator complex co –activator /co-repressor MED23 co –activation NCOR1 Genes ME with TTN are predictors of better survival 100% Cases with Alteration(s) in Query Gene(s) Cases without Alteration(s) in Query Gene(s) Cases with Alteration(s) in Query Gene(s) Logrank Test P-Value: 100%0.0283 Cases without Alteration(s) in Query Gene(s) 90% Logrank Test P-Value: 0.0425 90% 80% 80% 70% 70% 60% 60% Surviving 50% Disease Free 50% 40% 40% 30% 30% 20% 20% 10% 10% 0% 0% 0 20 40 60 80 100 120 140 160 180 200 220 240 0 20 40 60 80 100 120 140 160 180 200 220 240 Months Disease Free Months Survival 36 Summary • Introduction of mutual exclusivity classes and their relation to interaction network • Combining ME with interaction network improves identification of PanCancer dysregulated modules • Mutual exclusivity and co-occurrence of passenger mutations can provide important insights into mutagenesis of cancer 37 AlgoCSB Algorithmic Methods in Computational and Systems Biology Przytycka’s group Phung Dao Jan Hoinka YooAh Kim Damian Wojtowicz Yijie Wang DongYeon Cho Sanna Madan (alumnae) Poolesville HS .