Abstract 4332: Meta-analysis of molecular features associated with aggressive prostate cancer

Abstract

Abstract The TMPRSS2:ERG fusion is the most common genomic alteration in prostate cancer, occurring in 40-50% of primary tumors. The frequency of the ERG fusion remains between 40-50% in metastatic tumors, supporting previous findings that the fusion can drive tumor development, but is not sufficient to drive aggressive, lethal disease. Large -omic datasets collected over thousands of patient tumors, such as The Cancer Genome Atlas (TCGA), paired with outcomes present the raw data needed to associate genetic aberrations with markers of aggressive disease. Accordingly, we performed a meta-analysis of over 2,000 patient samples to identify combinations of -omic features associated with aggressive ERG fusion-positive and ERG fusion-negative prostate cancer. The data were collected from 8 independent, publicly available prostate cancer cohorts and cover range of disease phenotypes (e.g., neuroendocrine, castration resistance). Our analysis consisted of training univariate Cox regression models separately for single genomic or pairs of genomic features within each independent dataset. Based on the availability, we explored gene expression, mutation, and copy number alterations, all used to build separate models. We then combined models for a genomic feature across all datasets by weighting individual coefficients by the inverse of their squared error in a fixed effects model; final results are reported as corrected p-values. We compared and contrasted results from combined models within data type (i.e., only gene expression) to models integrating multiple data types. Interestingly, we find that the genomic features associated with aggressive disease differed based on ERG status. We report genomic loci that are altered in all patients, but associate with aggressive disease only in patients with an ERG fusion. We also report loci that co-occur or are mutually exclusive of the ERG fusion and associated with aggressiveness. For example, SPOP mutations are known to be mutually exclusive of ERG fusions. We report a finer stratification of patients with SPOP mutations based on biochemical relapse. Our results present a compressive meta-analysis of molecular features in prostate cancer and identify novel molecular subtypes associated with aggressive disease. Citation Format: James C. Costello, Rani Powers, Andrew Goodspeed, Scott D. Cramer. Meta-analysis of molecular features associated with aggressive prostate cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 4332.