Section: New Results
Convex-PL and entropy
During the 2019 we continued developing our knowledge-based scoring function for protein-ligand interactions called Convex-PL [44]. We introduced a new descriptor characterizing side-chain entropy, tried two ways of computing the solvent-accessible surface area (SASA) descriptors with either using buried SASA, or SASA of those atoms that are contributing to the protein-ligand interaction according to the scoring function design (i.e which are within the cutoff distance of an interaction potential). Using these descriptors alongwith Convex-PL score and ligand side chain entropy descriptor, we trained a ridge regression model to predict binding affinities. This modification of Convex-PL was assessed on the CASF Benchmark 2016 released in the end of 2018, and on a subset of the DUD benchmark [42]. We are currently preparing the corresponding manuscript for submission.