Benchmarking and biomarkers: the molecular and technical heterogeneity of prostate cancer

Paul Boutros1

1Ontario Institute for Cancer Research, Toronto, Canada


Prostate cancer is the leading non-skin malignancy in men, affecting 1 in 7 men over their lifetime. Today the vast majority of new diagnoses are of localised disease, and these patients have their treatment determined using the clinical/pathological factors of pre-treatment serum PSA levels, T-category and the Gleason Score. These clinical metrics are used to stratify men into low-, intermediate- and high-risk groups. My presentation will describe results from the Canadian Prostate Cancer Genome Network (CPC-GENE), the world's largest prostate cancer genomics project. CPC-GENE is sequencing the somatic and germline genomes of 500 intermediate-risk prostate cancers, along with epigenetic, transcriptomic and proteomic profiles. I will first discuss the intra-tumoural heterogeneity of these tumours, which shows profound genomic and micro-environmental heterogeneity. I will then link this variability directly to patient outcome, demonstrating the benefit of linking tumour hypoxia and somatic mutation profiles to robustly predict patient survival. Finally, I will discuss the challenges in translating sequencing-based assays into clinical practice. The ICGC-TCGA DREAM Somatic Mutation Calling Challenge provides the first rigorous investigation of the error profiles of sequencing studies, and I will discuss the startling degree of heterogeneity introduced by variability in analysis. Challenge-style benchmarking evaluations are increasingly critical, and I close by discussing how moving these into the cloud provides a framework for robust, rapid and reproducible comparison and optimization of algorithms for computational cancer biology.