Computer classification and construction of a novel prognostic signature based on moonlighting genes in prostate cancer

Abstract

Advanced prostate cancer (PRAD) patients have poor prognosis and rising morbidity despite the ongoing iteration of molecular therapeutic agents. As newly discovered proteins with several functions, Moonlighting proteins have showed an important role in tumor progression but has not been extensively investigated in PRAD. Our study aimed to identify moonlighting-related prognostic biomarkers and prospective PRAD therapy targets. 103 moonlighting genes were gathered from previous literatures. A PRAD classification and multivariate Cox prognostic signature were constructed using dataset from The Cancer Genome Atlas (TCGA). Subsequently, we tested our signature’s potential to predict biochemical failure-free survival (BFFS) using GSE21032, a prostate cancer dataset from Gene Expression Omnibus (GEO). The performance of this signature was demonstrated by Kaplan-Meier (KM), receiver operator characteristic (ROC), areas under ROC curve (AUC), and calibration curves. Additionally, immune infiltration investigation was conducted to determine the impact of these genes on immune system. This signature’s influence on drug susceptibility was examined using CellMiner’s drug database. Both training and validation cohorts demonstrated well predictive capacity of this 9-gene signature for PRAD. The 3-year AUCs for TCGA-PRAD and GSE21032 were 0.802 and 0.60 respectively. It can effectively classify patients into various biochemical recurrence risk groups. These genes were also assessed to be connected with tumor mutation burden (TMB), immune infiltration and therapy. This work created and validated a moonlighting gene signature, revealing fresh perspectives on moonlighting proteins in predicting prognosis and improving treatment of PRAD.