Prediction of Accurate Binding Modes using Combination of classical and accelerated Molecular dynamics and Free Energy Perturbation Calculations: An Application to Toxicity Studies
Micar21 Ltd., Persenk Str. 34B, 1407 Sofia, Bulgaria
Received: ...2018 ; Accepted: .... 2018; Accepted author version posted online: ...2018
Estimating the correct binding modes of ligands in protein-ligand complexes is not only crucial in the drug discovery process, but also for elucidating potential toxicity mechanisms. In the current paper, we discuss and demonstrate a computational modelling protocol using the combination of docking, classical (cMD) and accelerated (aMD) molecular dynamics and free energy perturbation (FEP+ protocol) for identification of the binding modes of selected perfluorocarboxyl acids (PFCAs) in the PPARγ nuclear receptor. Initially, we employed both the regular and induced fit docking which failed to correctly predict the ligand binding modes and rank the compounds with respect to experimental free energies of binding, when they were docked into non-native X-ray structure. The cMD and aMD simulations identified the presence of multiple binding modes for these compounds, and the shorter chain PFCAs (C6-C8) continuously moved between a few energetically favourable binding conformations. These results demonstrate that the docking scoring function cannot rank compounds precisely in such cases, not due to its insufficiency, but because of the use of incorrect or only one unique bindings pose, neglecting the protein dynamics. Finally, based on MD predictions of binding conformations, the FEP+ sampling protocol was extended and then accurately reproduced experimental differences in the free energies. Thus, the preliminary MD simulations can also provide helpful information about correct set-up of the FEP+ calculations. These results show that the PFCAs binding modes were accurately predicted and the FEP+ protocol can be used to estimate free energies of binding of flexible molecules outside of typical drug-like compounds. Our in silico workflow revealed the main characteristics of the PFCAs, which are week PPARγ partial agonists and illustrated the importance of specific ligand-residue interactions within the LBD. This work also suggests a common workflow for identification of ligand binding modes, ligand-protein dynamics description and relative free energy calculations.