Two-sample Mendelian randomization combined with GraphBAN reveals druggable prognostic genes and risk stratification in kidney renal clear cell carcinoma

Purpose: Kidney renal clear cell carcinoma (KIRC) remains associated with poor outcomes in advanced disease, and robust prognostic biomarkers and actionable therapeutic targets are still lacking. This study aimed to identify druggable genes causally associated with KIRC and to establish a clinically relevant risk stratification model. Methods: We integrated TCGA-KIRC transcriptomic data, FinnGen GWAS summary statistics, curated druggable gene sets, differential expression analysis, two-sample Mendelian randomization, Cox and LASSO modeling, immune and mutation profiling, and GraphBAN-based compound prediction with molecular docking and molecular dynamics validation. Results: We identified 129 druggable genes with significant causal associations with KIRC and established a four-gene prognostic signature comprising ALDH6A1, AMH, ASIC3, and PLK1. The resulting risk model effectively stratified patients into high- and low-risk groups, with poorer survival in the high-risk group and good predictive performance in both internal and external cohorts (AUC > 0.70). High-risk tumors also showed distinct pathway enrichment, an altered immune microenvironment, lower predicted immunotherapy response, and a higher somatic mutation burden. In addition, several candidate compounds showed favorable binding to prognostic targets. Conclusions: These findings identify clinically relevant druggable prognostic genes in KIRC and provide a useful framework for risk assessment, target discovery, and individualized therapeutic exploration.