Survival Prediction for Prostate Cancer Using Machine Learning and Bioinformatics Models

Prostate cancer (PC) is the most commonly diagnosed and the second most lethal malignancy in men. Proper understanding about the factors influencing the disease mechanism, response to the treatment and long term survival could facilitate effective disease management, treatment planning and decision making. Previous research initiatives reported a number of genes having impact on PC development but their genetic influence on the overall survival of the patients is still obscure. In this study, we fist identified PC related signature genes by analysing the RNA-seq transcriptomic data. Then we investigated the influence of those genes on the survival of PC patients using the clinical and transcriptomic data from the Cancer Genome Atlas (TCGA). Considering the univariate and multivariate analysis using the Cox proportional-hazards (CoxPH) model, we evidenced notable variation in the survival period between the altered and normal groups for two genes (APLN, and DUOXA1). We also identified ten hub genes such as RRAS, TUBB4A, CAV1, EFNB1, RHOU, ITGB4, TPM2, MYL9, UBC, and TGFB3 in protein-protein interaction analysis that could be the source of potential therapeutic intervention. Moreover, several significant molecular pathways through functional enrichment analysis was obtained. After verification through functional studies, the identified genetic determinants could serve as therapeutic target for prolonged PC survival.