Precision medicine insight into primary prostate tumor through transcriptomic data and an integrated systems biology approach

Abstract The non-responsiveness towards the therapy of prostate tumors is due to an extreme genetic heterogeneity including various tumor subpopulations depending on the genetic variations underlying this disease. However, the determination of prostate tumor clinical subpopulations, important for therapeutic decisions, is challenging. It is also key step in the precision medicine approaches. To facilitate a quick and accurate determination of clinical subpopulations with potential therapeutic targets, we developed a workflow using a two-step pipeline. This includes 1) an unsupervised machine learning classifier, based on consensus clustering to identify primary prostate tumor subpopulations; and 2) Weighted Gene Co-expression Network Analysis (WGCNA) to predict key molecular signatures in each of the tumor subpopulations. In a case study on genomic profiling of human prostate cancer data, available in public domain, we found four prostate tumor subpopulations, each with distinct set of gene expression patterns as well as clinical characteristics that can be investigated for potential therapeutic interventions. Furthermore, we analyzed patient derived biochemical recurrence (BCR) data using the Kaplan-Meier survival analysis and found a significant correlation between the signatures identified and the clinical phenotype observed. We believe that the explored signatures for the identified primary prostate tumor subpopulations in this study can be considered as potential diagnostic and therapeutic targets for development of precision medicine approach for primary prostate tumor.