Development of a Prognostic Index for Clear Cell Renal Cell Carcinoma Using Machine Learning and Multi-omics Analysis to Enhance Clinical Outcomes and Drug Sensitivity

Background Clear cell renal cell carcinoma (KIRC) represents the major malignant subtype of kidney cancer and shows pronounced biological heterogeneity. For patients with advanced disease, therapeutic resistance and the lack of stable prognostic markers continue to keep 5-year survival below 15%. Cancer driver genes (CDGs) participate in malignant transformation and tumor progression, but their combined prognostic value in KIRC has not been fully clarified. Methods We assembled a catalogue of 6,291 CDGs and designed a machine-learning-based workflow to build a Cancer Driver Gene Prognostic Index (CDPI). The analysis integrated differential expression screening, consensus clustering, WGCNA, and four feature-selection algorithms, including Lasso, decision tree, random forest, and XGBoost. Model performance was examined in the TCGA-KIRC training set and then tested in the external E-MTAB-1980 cohort. Results Differential expression screening yielded 5,243 DECDRGs and four recurrent core DECDRGs. Based on consensus clustering, KIRC samples were separated into two molecular groups, and the C1 group showed poorer survival. WGCNA selected the turquoise module as the cluster-associated module, from which 148 key DECDRGs were obtained; 105 of these genes were significant in univariate Cox analysis. Cross-model feature selection ultimately produced a CDPI composed of five genes: PLCL1, GABRB3, USP46, RNF152, and PFKP. Patients with higher CDPI values had shorter overall survival in both datasets. The CDPI-based nomogram achieved good prediction accuracy, including a 1-year AUC of 0.86, and showed a favorable decision-curve profile. High CDPI was also accompanied by stronger immune infiltration, lower tumor purity, and different mutation patterns. Drug-response prediction suggested increased sensitivity to gemcitabine, epirubicin, ULK1 inhibitors, docetaxel, and AZD7762, but reduced sensitivity to Daporinad, osimertinib, and cediranib. Conclusion The CDPI may serve as a practical stratification index for KIRC and may help interpret immune escape and treatment vulnerability, although prospective clinical confirmation is still required.