A Promising Prognostic Signature Consisting of Fatty Acid Metabolism Genes based on Machine Learning Predicts Biochemical Recurrence and Aids ARSI Therapy in Prostate Cancer

Background: Fatty acid metabolism (FAM) is a crucial metabolic characteristic of tumor cells, playing a role in various pathological processes during tumor development.Till now, the prognostic role of FAM-related genes of prostate cancer (PCa) is far from fully investigation.Methods: The combinations of 10 machine learning algorithms were applied in this study.A reliable signature, FAM-related gene score (FAMRGs), was developed to predict the prognosis of patients with PCa.External data sets were used to verify the accuracy and robustness of the FAMRGs.Drug sensitivity analysis was used to predict the optimal drug for high-risk PCa patients.The underlying mechanism related to FAMRGs were investigated by functional enrichment analysis.A nomogram based on FAMRGs was developed for personalized prediction of patient prognosis.Results: A stable FAMRGs was construced and validated in 6 independent cohorts.FAMRGs accurately divided PCa patients into low and high risk group.FAMRGs showed stronger predictive ability compared with published prognostic signatures for PCa.Also, the androgen receptor signaling inhibitors (ARSI) treatment response predictive ability of FAMRGs was identified.Five drugs that were most suitable for patients in the high risk group of FAMRGs were screened.It was shown that FAMRGs involved in cell cycle-related pathways.The novel nomogram showed precisely predictive ability for the outcomes of patients with PCa.Conclusions: The FAMRGs can accurately predict the prognosis of PCa patients and is expected to direct the clinical treatment for PCa.