Abstract
Elucidating the correct binding mode of drug-like compounds is crucial to reveal the molecular determinants underlying the recognition by the target protein and to estimate the binding affinity, thus guiding ensuing hit-to-lead optimization. However, identifying the near-native binding pose in docking experiments remains a major hurdle. Assuming that the hydropathic complementarity principle is the major driving force in ligand-protein recognition, the suitability of simple hydropathicity-based scoring functions to reveal the near-native pose was examined. A benchmarking dataset of 1000 ligand-protein complexes purposely designed to encompass bioactive and decoy poses was used to assess the performance of 3D hydropathicity atomic descriptors derived from empirical and quantum mechanical models. This strategy led to a predictive accuracy of ca. 90% when the non-negligible influence of the conformational stress is also considered. The results reveal the need to leverage an understanding of nonpolar/polar chemical features for the successful discrimination of the near-native pose.