Networks of biological molecules are key to cellular function, governing processes ranging from signal cascade propagation to metabolic pathway regulation. Genetic duplication processes give rise to sets of regulatory proteins that have evolved from a common ancestor, leading to interactomes whose dysregulation is often associated with disease. A better understanding of the determinants of specificity at 21 interfaces shared by functionally related proteins is crucial to the rational design of novel pharmacotherapeutic agents. To this end, a comprehensive dataset of drug and drug-like binders was assembled for the Bcl-xL and Bcl-2 antiapoptotic proteins – archetypal examples of regulatory systems governed by evolutionarily conserved protein-protein interactions. These were first used to derive a two-dimensional quantitative structure-activity relationship (2D QSAR) model, predicting ligand specificity for these homologous proteins. The strengths and weaknesses of high-throughput 2D QSAR were then compared and contrasted to those of theoretically rigorous thermodynamic integration calculations performed on 14 complexes of Bcl-xL-specific, Bcl-2-specific, and potent dual binders bound to the Bcl-xL and Bcl-2 proteins. We demonstrate that free energy calculations provide an added layer of essential information, which traditional QSAR cannot capture. Moreover, we show that protein energetic responses to different ligands, expressed as per-residue energy values, can be used to fingerprint the protein-ligand interaction, extending the framework of four-dimensional molecular dynamics/quantitative structure-activity relationships (4D-MD/QSAR) towards the facilitation of future drug design strategies.
Journal of Chemical Inofrmation and Modeling, 2018 Dec 24. doi: 10.1021/acs.jcim.8b00765