Sergei A. Nikolenko

This study benchmarks 11 protein-ligand interaction prediction tools, showing classical and deep learning methods perform similarly on drug-like ligands, while co-folding tools excel on complex ligands and all struggle with generalization to unseen pockets.