Discovery of mitochondrial modulators using target-aggregated machine learning

• A target-agnostic, data-integrated machine learning strategy was developed to overcome data scarcity and enable robust predictive modeling. • Multiple machine learning and deep learning algorithms were systematically evaluated, with XGBoost achieving a top ROC AUC of 0.943. • Virtual screening of a large chemical library identified six novel mitochondrial modulators, all experimentally validated in human cell assays. • A scalable, target-independent framework was established for phenotypic drug discovery in complex biological systems.