Autophagy Associated Genes (ARGs) ‐Based Predictive Model AIDPS for Prostate Cancer

ABSTRACT Prostate cancer (PCa) is one of the most common cancers in men worldwide. Autophagy‐related genes (ARGs) may play an important role in various biological processes of PCa. The aim of this study was to identify and evaluate autophagy‐related features to predict clinical outcomes in patients with PCa. Single‐cell sequencing data and RNA sequencing data was included from public GEO and TCGA databases. Cells were clustered and annotated by dimension reduction cluster analysis. Epithelial cells, T cells and fibroblasts were isolated to explore their heterogeneity. Autophagy‐related genes were obtained from the HADb database. Survival analysis was conducted by K–M curve, and prognostic risk model was established using 101 machine learning algorithms. In addition, we performed gene colocalisation analysis and Mendelian randomisation analysis. Univariate Cox analysis was used to screen out prognostic genes from DEGs and ARGs in each dataset. Risk model was generated by artificial intelligence‐derived prognostic signature (AIDPS), which showed better prognostic performance in every dataset than other published models for PCa. The disease‐free period (DFS) of patients in the high‐risk group was significantly worse than that in the low‐risk group (all p < 0.05). The best model is the Ridge (C‐index 0.726). We found significant differences in IC50 values of the Dactinomycin_1811, Dactolisib_1057, Luminespib_1559 and Paclitaxel_1080 between groups. In SNP sites rs2743987 and rs7768988, there was a significant correlation between prostate hyperplasia and prostate cancer. Our ARG‐based predictive model AIDPS is a reliable and effective tool for prognosis and treatment of prostate cancer.