Drug interaction profiles classification: Sørensen–Dice coefficient based on MM/GBSA energies to explore focal adhesion kinase inhibitors

Abstract

Quantitative comparison of ligand-target interaction profiles is often challenging due to subjective interpretations in visual inspection or limitation to predefined interactions in analytical methods. Moreover, traditional analyses tend to overlook the dynamic nature of the interactions, focusing instead on mean energy values. By describing ligand-residue interactions as energy distributions, we account for these dynamics. Assuming a Gaussian distribution, the mean and standard deviation are sampled from molecular dynamics simulations. Using the Sørensen-Dice similarity index, we constructed a metric where the overlap between Gaussian interactions quantifies their similarity. For ligand comparison, the average per-residue similarity is employed. This method allows for post-processing techniques like clustering and dimensionality reduction. We applied it to the focal adhesion kinase (FAK) type II inhibitors targeting the ATP binding site, identifying interaction patterns among the inhibitors.