Abstract
We report the implementation of molecular modeling approaches developed as a part of the 2016 Grand Challenge 2, the blinded competition of computer aided drug design technologies held by the D3R Drug Design Data Resource (https://drugdesigndata.org/). The challenge was focused on the ligands of the farnesoid X receptor (FXR), a highly flexible nuclear receptor of the cholesterol derivative chenodeoxycholic acid. FXR is considered an important therapeutic target for metabolic, inflammatory, bowel and obesity related diseases (Expert Opin Drug Metab Toxicol 4:523-532, 2015), but in the context of this competition it is also interesting due to the significant ligand-induced conformational changes displayed by the protein. To deal with these conformational changes we employed multiple simulations of molecular dynamics (MD). Our MD-based protocols were top-ranked in estimating the free energy of binding of the ligands and FXR protein. Our approach was ranked second in the prediction of the binding poses where we also combined MD with molecular docking and artificial neural networks. Our approach showed mediocre results for high-throughput scoring of interactions.
Highlights
The 2016 drug design Grand Challenge 2 competition consisted of six blind sub-challenges focused on the farnesoid X receptor (FXR) protein: (a) the ligand-based prediction of binding potency for 102 compounds with an experimentally measured affinity, (b) the prediction of spatial coordinates of 36 molecular complexes with experimentally resolved atomic coordinates, (c) the high-throughput prediction of binding potency by structure based methods for the 102 compounds and (d) relative free energy estimations within the subset of homologous compounds
Our approach was prototyped and described previously [25] and here we have further extended it with an formal assessment of potential of mean force (PMF) convergence for the tracking of the estimation outliers
In all our protocols described molecular dynamic simulations are combined with variety of other methods
Summary
The 2016 drug design Grand Challenge 2 competition consisted of six blind sub-challenges focused on the FXR protein: (a) the ligand-based prediction of binding potency for 102 compounds with an experimentally measured affinity, (b) the prediction of spatial coordinates of 36 molecular complexes with experimentally resolved atomic coordinates (pose prediction), (c) the high-throughput prediction of binding potency by structure based methods (before and after disclosure of the 36 experimentally resolved complexes) for the 102 compounds and (d) relative free energy estimations within the subset of homologous compounds (two subsets of 15 sulfonamides and 18 spiros compounds). The goal of structure-based high throughput virtual screening is to find and evaluate energetically favorable binding modes between a target protein and millions of candidate small organic compounds in a timeframe for which one is willing to tolerate. While it is believed that docking algorithms perform reasonably well in determining the geometries of the docked complexes, they usually fail to accurately evaluate the corresponding free energy of binding [2, 3]. This may not be surprising since even an approximate estimation of entropy—the major term describing free energy in biomolecular complexes—requires
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
