Machine Learning-based prediction of protein-ligand binding affinity

Machine Learning-based prediction of protein-ligand binding affinity F. Tandaa, R. Beccariab,c,d, M. Civeraa, G. Tianab,e a Università degli studi di Milano, Dipartimento di Chimica, via Golgi 19, 20133, Milano, Italia. b Università degli studi di Milano, Dipartimento di Fisica, via Celoria 16, 20133, Milano, Italia. c Heidelberg Institute for Theoretical Studies, Schloss-Wolfsbrunnenweg 35, 69118, Heidelberg, Germania d Faculty of Physics, Heidelberg University, Im Neuenheimer Feld 227, 69120 Heidelberg, Germania e INFN, via Celoria 16, 20133, Milano, Italia E-mail: [email protected] Accurately predicting ligand-protein binding affinity is crucial in computational drug discovery, as it enables the efficient identification and optimization of therapeutic compounds. Machine learning can capture complex molecular interaction patterns beyond traditional methods, yet existing models are often constrained by dataset quality and molecular representation [1,2]. We propose a two-stage model for compound ranking, in which a classifier first identifies likely binders and a regressor subsequently predicts binding constants used to rank compounds. Our approach achieves predictive performance comparable to state-of-the-art models, while substantially improving computational efficiency, enabling the screening of tens of thousands of compounds within minutes. References [1] Gaillard T., Journal of Chemical Information and Modeling 2018, 58(8):1697–1706 [doi:10.1021/acs.jcim.8b00312] [2] Wang H., Briefings in Bioinformatics 2024, 25(2):bbae081 [doi: 10.1093/bib/bbae081]