On importance of explicit account of non-complementary contacts in scoring functions

Based both on the practice of post-processing by a human expert and on the higher values of the accuracy metrics of machine learning scoring functions, it is suggested that when estimating the free energy of binding in a ligand–receptor complex, a significant part of intermolecular interactions is still not explicitly taken into account. An assessment is made of how explicit consideration of non-complementary ligand–receptor interactions could improve the accuracy of the description of contemporary classical scoring functions, which tend to use only terms of complementary/favorable interactions.