Structure-based virtual screenings (SBVSs) play an important role in drug discovery projects. However, it is still a challenge to accurately predict the binding affinity of an arbitrary molecule binding to a drug target and prioritize top ligands from a SBVS. In this study, we combined ligand-residue interaction profiles (IPs) and machine learning (ML) as a novel method to construct prediction models, which significantly improved the screening performance in SBVSs. The resulting prediction models are called IP scoring functions (IP-SFs). We systematically investigated how to improve the performance of IP-SFs from many perspectives, including different sampling methods before interaction energy calculation and different ML algorithms. Using six drug targets with each having hundreds of known ligands, we conducted a critical evaluation on the developed IP-SFs. The IP-SFs employing a gradient boosting decision tree (GBDT) algorithm in conjunction with the MIN+GB simulation protocol achieved the best overall performance. Its scoring power, ranking power and screening power significantly outperformed the Glide SF. First, compared to Glide, the average values of mean absolute error and root mean square error of GBDT/MIN+GB decreased by about 38% and 36%, respectively. Second, the mean values of squared correlation coefficient and predictive index increased by about 225% and 73%, respectively. Third, more encouragingly, the average value of the areas under the curve of receiver operating characteristic (ROC) for six targets by GBDT was 0.87, which was significantly better than that by Glide, which was only 0.71. Therefore, we expect IP-SFs to have broad and promising applications in SBVSs.

Machine learning on ligand-residue interaction profiles to significantly improve binding affinity prediction

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.