Cancer is driven by genetic change, and the advent of massively parallel sequencing has enabled systematic documentation of this variation at the whole-genome scale^1–3. Here we report the integrative analysis of 2,658 whole-cancer genomes and their matching normal tissues across 38 tumour types from the Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium of the International Cancer Genome Consortium (ICGC) and The Cancer Genome Atlas (TCGA). We describe the generation of the PCAWG resource, facilitated by international data sharing using compute clouds. On average, cancer genomes contained 4–5 driver mutations when combining coding and non-coding genomic elements; however, in around 5% of cases no drivers were identified, suggesting that cancer driver discovery is not yet complete. Chromothripsis, in which many clustered structural variants arise in a single catastrophic event, is frequently an early event in tumour evolution; in acral melanoma, for example, these events precede most somatic point mutations and affect several cancer-associated genes simultaneously. Cancers with abnormal telomere maintenance often originate from tissues with low replicative activity and show several mechanisms of preventing telomere attrition to critical levels. Common and rare germline variants affect patterns of somatic mutation, including point mutations, structural variants and somatic retrotransposition. A collection of papers from the PCAWG Consortium describes non-coding mutations that drive cancer beyond those in the TERT promoter⁴; identifies new signatures of mutational processes that cause base substitutions, small insertions and deletions and structural variation^5,6; analyses timings and patterns of tumour evolution⁷; describes the diverse transcriptional consequences of somatic mutation on splicing, expression levels, fusion genes and promoter activity^8,9; and evaluates a range of more-specialized features of cancer genomes^8,10–18.

Malignant peritoneal mesothelioma (PeM) is a rare and fatal cancer that originates from the peritoneal lining of the abdomen. Standard treatment of PeM is limited to cytoreductive surgery and/or chemotherapy, and no effective targeted therapies for PeM exist. Some immune checkpoint inhibitor studies of mesothelioma have found positivity to be associated with a worse prognosis. To search for novel therapeutic targets for PeM, we performed a comprehensive integrative multi-omics analysis of the genome, transcriptome, and proteome of 19 treatment-naïve PeM, and in particular, we examined BAP1 mutation and copy number status and its relationship to immune checkpoint inhibitor activation. We found that PeM could be divided into tumors with an inflammatory tumor microenvironment and those without and that this distinction correlated with haploinsufficiency of BAP1. To further investigate the role of BAP1, we used our recently developed cancer driver gene prioritization algorithm, HIT’nDRIVE, and observed that PeM with BAP1 haploinsufficiency form a distinct molecular subtype characterized by distinct gene expression patterns of chromatin remodeling, DNA repair pathways, and immune checkpoint receptor activation. We demonstrate that this subtype is correlated with an inflammatory tumor microenvironment and thus is a candidate for immune checkpoint blockade therapies. Our findings reveal BAP1 to be a potential, easily trackable prognostic and predictive biomarker for PeM immunotherapy that refines PeM disease classification. BAP1 stratification may improve drug response rates in ongoing phases I and II clinical trials exploring the use of immune checkpoint blockade therapies in PeM in which BAP1 status is not considered. This integrated molecular characterization provides a comprehensive foundation for improved management of a subset of PeM patients.

The catalog of cancer driver mutations in protein-coding genes has greatly expanded in the past decade. However, non-coding cancer driver mutations are less well-characterized and only a handful of recurrent non-coding mutations, most notably TERT promoter mutations, have been reported. Here, as part of the ICGC/TCGA Pan-Cancer Analysis of Whole Genomes (PCAWG) Consortium, which aggregated whole genome sequencing data from 2658 cancer across 38 tumor types, we perform multi-faceted pathway and network analyses of non-coding mutations across 2583 whole cancer genomes from 27 tumor types compiled by the ICGC/TCGA PCAWG project that was motivated by the success of pathway and network analyses in prioritizing rare mutations in protein-coding genes. While few non-coding genomic elements are recurrently mutated in this cohort, we identify 93 genes harboring non-coding mutations that cluster into several modules of interacting proteins. Among these are promoter mutations associated with reduced mRNA expression in TP53, TLE4, and TCF4. We find that biological processes had variable proportions of coding and non-coding mutations, with chromatin remodeling and proliferation pathways altered primarily by coding mutations, while developmental pathways, including Wnt and Notch, altered by both coding and non-coding mutations. RNA splicing is primarily altered by non-coding mutations in this cohort, and samples containing non-coding mutations in well-known RNA splicing factors exhibit similar gene expression signatures as samples with coding mutations in these genes. These analyses contribute a new repertoire of possible cancer genes and mechanisms that are altered by non-coding mutations and offer insights into additional cancer vulnerabilities that can be investigated for potential therapeutic treatments.

Systemic targeted therapy in prostate cancer is primarily focused on ablating androgen signaling. Androgen deprivation therapy and second-generation androgen receptor (AR)–targeted therapy selectively favor the development of treatment-resistant subtypes of metastatic castration-resistant prostate cancer (mCRPC), defined by AR and neuroendocrine (NE) markers. Molecular drivers of double-negative (AR−/NE−) mCRPC are poorly defined. In this study, we comprehensively characterized treatment-emergent mCRPC by integrating matched RNA sequencing, whole-genome sequencing, and whole-genome bisulfite sequencing from 210 tumors. AR−/NE− tumors were clinically and molecularly distinct from other mCRPC subtypes, with the shortest survival, amplification of the chromatin remodeler CHD7, and PTEN loss. Methylation changes in CHD7 candidate enhancers were linked to elevated CHD7 expression in AR−/NE+ tumors. Genome-wide methylation analysis nominated Krüppel-like factor 5 (KLF5) as a driver of the AR−/NE− phenotype, and KLF5 activity was linked to RB1 loss. These observations reveal the aggressiveness of AR−/NE− mCRPC and could facilitate the identification of therapeutic targets in this highly aggressive disease. Significance: Comprehensive characterization of the five subtypes of metastatic castration-resistant prostate cancer identified transcription factors that drive each subtype and showed that the double-negative subtype has the worst prognosis.

Machine learning methods are used to discover complex nonlinear relationships in biological and medical data. However, sophisticated learning models are computationally unfeasible for data with millions of features. Here we introduce the first feature selection method for nonlinear learning problems that can scale up to large, ultra-high dimensional biological data. More specifically, we scale up the novel Hilbert-Schmidt Independence Criterion Lasso (HSIC Lasso) to handle millions of features with tens of thousand samples. The proposed method is guaranteed to find an optimal subset of maximally predictive features with minimal redundancy, yielding higher predictive power and improved interpretability. Its effectiveness is demonstrated through applications to classify phenotypes based on module expression in human prostate cancer patients and to detect enzymes among protein structures. We achieve high accuracy with as few as 20 out of one million features --- a dimensionality reduction of 99.998%. Our algorithm can be implemented on commodity cloud computing platforms. The dramatic reduction of features may lead to the ubiquitous deployment of sophisticated prediction models in mobile health care applications.

Prioritizing molecular alterations that act as drivers of cancer remains a crucial bottleneck in therapeutic development. Here we introduce HIT'nDRIVE, a computational method that integrates genomic and transcriptomic data to identify a set of patient-specific, sequence-altered genes, with sufficient collective influence over dysregulated transcripts. HIT'nDRIVE aims to solve the “random walk facility location” (RWFL) problem in a gene (or protein) interaction network, which differs from the standard facility location problem by its use of an alternative distance measure: “multihitting time,” the expected length of the shortest random walk from any one of the set of sequence-altered genes to an expression-altered target gene. When applied to 2200 tumors from four major cancer types, HIT'nDRIVE revealed many potentially clinically actionable driver genes. We also demonstrated that it is possible to perform accurate phenotype prediction for tumor samples by only using HIT'nDRIVE-seeded driver gene modules from gene interaction networks. In addition, we identified a number of breast cancer subtype-specific driver modules that are associated with patients’ survival outcome. Furthermore, HIT'nDRIVE, when applied to a large panel of pan-cancer cell lines, accurately predicted drug efficacy using the driver genes and their seeded gene modules. Overall, HIT'nDRIVE may help clinicians contextualize massive multiomics data in therapeutic decision making, enabling widespread implementation of precision oncology.