Predicting prostate cancer clinical outcomes with germline genomic biomarkers

PROJECT SUMMARY/ABSTRACT Prostate cancer is the second-most diagnosed cancer and the second leading cause of cancer death in American men. Early detection of localized disease is common but is followed by the more challenging task of prognosing a highly variable clinical course. Localized tumors are often indolent, but not always. Current clinical risk-

assessment schemes incorporating serum abundance of prostate specific antigen (PSA), tumor size & extent, and tumor grade based on biopsy are limited in accuracy; over a quarter of patients are over-treated. The goal of this proposal is to fill the urgent need for better markers discriminating indolent from aggressive localized

prostate cancers. An improved method of risk stratification may lie in hereditary factors. Prostate cancer is one of the most strongly inherited (h2 = 57%), with accumulating evidence associating rare variants, common variants and genetic ancestry with clinical outcomes. The overall hypothesis of this proposal is that variant status of DNA

damage repair (DDR) genes, prostate cancer polygenic risk and genetic ancestry will improve clinical risk stratification of prostate cancer. To address this hypothesis, the aims of this proposal focus on two clinical outcomes representing key decision points in the management of the disease. Specifically, the aims are to

investigate the biomarker potential of the described germline factors to predict (1) progression from active surveillance of a patient, informing whether they should defer treatment, and (2) biochemical relapse after therapy of a patient, informing the most appropriate treatment course. For both aims, germline DNA samples

from a cohort of patients diagnosed with localized prostate cancer and with extensive clinical follow-up data will be sequenced using a targeted sequencing approach. A bespoke bioinformatics workflow will be used to derive germline genomic endpoints: pathogenic and likely pathogenic DDR variants, genomic risk score and genetic

ancestry. For each aim, exploratory univariate and multivariate analyses will validate associations between genomic and clinical endpoints, and control for clinical prognostic features. Multivariate Cox proportional-hazards regression models will be trained, validated, and evaluated for predictive performance to establish the utility of

germline genomic biomarkers in prostate cancer clinical risk stratification. The results of this proposal have the potential to transform standard-of-care management for this disease and thus substantially contribute to enhancing human health and quality of life.