Integrated machine learning algorithms for prediction of prognosis in ovarian cancer patients based on mitochondrial-related genes.

Mitochondrial dysfunction drives ovarian cancer (OC) progression. This study constructed a robust prognostic model (MITO-OC) based on mitochondria-related genes using ten machine-learning algorithms on TCGA, ICGC, and GEO data. We identified 241 differentially expressed genes and built the optimal MITO-OC model using StepCox[forward] and RSF algorithms (C-index=0.73). The model accurately predicts patient overall survival and strongly correlates with tumor immune infiltration. Furthermore, single-cell and pan-cancer analyses highlighted CHCHD2 as a critical component in OC and other tumors. MITO-OC provides a highly effective, personalized prognostic tool and reveals underlying metabolic mechanisms for OC clinical management.