AI and the World's Hard Problems · Medicine & drug discovery · Nature (computational biology) · 2026-10-25
Identification of a natural product derived NaV1.5 sodium channel modulator through integrated machine learning and molecular modeling
Existing sodium channel blockers carry proarrhythmic risks, so safer NaV1.5 modulators are needed for atrial fibrillation. The team trained LightGBM, XGBoost and Random Forest models on 1,512 known NaV1.5 inhibitors and screened 100,404 natural compounds from COCONUT. The best model (LightGBM, MCC 0.66) led to Piperaduncin C, which showed favorable docking, stable simulations and a predicted binding energy better than quinidine, though experimental validation is still required.