Section: New Results
Development of Convex-PL, a scoring function for protein-ligand interactions
Participants : Sergei Grudinin, Maria Kadukova.
We have continued developing Convex-PL, which is a knowledge-based scoring function for protein-ligand interactions. It is based on the assumption that protein-ligand interactions linearly depend on radial distributions of pairs of atoms of various types. The corresponding coefficients are deduced with a convex optimization problem from the structural data. We augmented Convex-PL with a term standing for steric clashes, ran a leave-one-out cross-validation procedure, and additionally validated it on the D3R Grand Challenge 2 user-submitted poses. The corresponding paper [8] was accepted to the Journal of Computer-Aided Molecular Design. We have also presented Convex-PL at two scientific conferences [53], [52].