Machine learning-based screening of potential diagnostic markers for prostate cancer with immune cell infiltration...

Prostate adenocarcinoma (PRAD) poses a significant global health burden. Apolipoprotein B mRNA editing enzyme catalytic polypeptide-like 3C (APOBEC3C) exhibits context-dependent roles in cancer but its function in PRAD remains unclear. This study aims to resolve APOBEC3C's functional duality in PRAD and identify immune-related hub genes using integrated bioinformatics and experimental validation.

Screening of PRAD differentially expressed genes (DEGs), Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and CIBERSORT immune infiltration assessment were performed based on The Cancer Genome Atlas (TCGA) database. Second, key genes in DEGs were screened using random forest (RF) analysis, the least absolute shrinkage and selection operator (LASSO) regression model, and the Support Vector Machine-Recursive Feature Elimination (SVM-RFE). The diagnostic value of the characterized core gene was evaluated by receiver operating characteristic (ROC) curve analysis in both TCGA and external Gene Expression Omnibus (GEO)-GSE29079 cohorts. The correlation between the core gene and tumor-infiltrating immune cells was further analyzed. In addition, in vitro PRAD models (PC-3/DU145 cell lines) were established to validate bioinformatics findings through functional assays, including quantitative real-time polymerase chain reaction (qRT-PCR), small interfering RNA (siRNA) knockdown, proliferation/migration tests, and chemokine profiling.

DEG analysis identified 24,208 genes (2,269 up-regulated and 1,966 down-regulated). GO enrichment analysis showed that DEGs were mainly involved in biological processes such as regulation of vascular contraction, maintenance of epithelial homeostasis in the gastrointestinal tract, and intercellular adhesion; the KEGG pathway was enriched in vascular smooth muscle contraction, neuroactive ligand-receptor interactions, and cyclic adenosine monophosphate (cAMP) signaling pathways. Immune infiltration showed distinct distribution of 21 immune cells in PRAD vs. normal tissues (P=0.045). Machine learning (LASSO/RF/SVM-RFE) identified APOBEC3C as a core gene with high diagnostic accuracy [TCGA area under the curve (AUC) =0.859; GEO AUC =0.835]. Further analysis showed that APOBEC3C expression was significantly positively correlated with anti-tumor immune cells (activated dendritic cells: r=0.45, P=0.008; resting memory CD4⁺ T cells: r=0.38, P=0.015). The qRT-PCR results showed that APOBEC3C was expressed at significantly lower mRNA levels in PRAD cell lines (PC-3: 0.32±0.05-fold, P=0.003; DU145: 0.28±0.04-fold, P=0.001) than in normal prostate epithelial cells (RWPE-1). Functional validation confirmed that siRNA-mediated APOBEC3C knockdown (>75% efficiency) promoted malignant phenotypes: proliferation increased 1.8-fold (P<0.001) and migration accelerated 2.1-fold (P<0.001). Crucially, knockdown suppressed T-cell-recruiting chemokines C-X-C motif chemokine ligands 9 (CXCL9) (62% reduction) and C-X-C motif chemokine ligands 10 (CXCL10) (60% reduction) in PRAD cells (P<0.001).

This study revealed the regulatory role of APOBEC3C as an oncogene in PRAD progression, suggesting that it may become a potential molecular marker for PRAD diagnosis and targeted therapy and providing new ideas for clinical intervention.