Prediction of the PPARα agonism of fibrates by combined MM–docking approaches

Fibrates are peroxisome proliferator-activated alpha receptor (PPARα) activators derived from fibric acid and are the most clinically used therapeutics in the treatment of hypertriglyceridemia. Long standing studies on these drugs have accumulated a large body of experimental data about their biological activity and, more recently, on the molecular mechanism mediating their PPARα agonism. An immense interest for the discovery of new fibrates with improved potency and PPARα selectivity has stimulated many investigations toward a deeper understanding of structure–activity relationships controlling their activity. The present study aimed at investigating the binding properties of a set of 23 fibrates, characterized by similar carboxylic heads but differing in the size and orientation of the hydrophobic portion, using computational approaches. We combined standard docking and molecular mechanics approaches to better describe the adaptation of the protein target to the bound ligand. The agonist potencies were then regressed against the calculated binding energies to elaborate predictive model equations. The obtained models were characterized by good performances realizing a fair trade-off between accuracy and computational costs. The best model was obtained with a regression procedure allowing automatic generation of a training subset from the whole set of trials and filtering out outliers, thus highlighting the importance of regression strategies.

Figure optionsDownload high-quality image (222 K)Download as PowerPoint slideHighlights
► The present investigation focused on the estimation of the binding geometries and the molecular interactions of a series of 23 fibrates with the PPARα receptor through a LIECE-like approach based on the combination of docking and MM calculations.
► The receptor-fibrate affinity was estimated as the energy difference between the optimized complex and the separated receptor and ligand structures calculated in a continuous solvent simulating the water solution (solvation docking).
► Regression analyses of the experimental agonist potencies with respect to the theoretical estimations of target-ligand binding energies were performed to derive predictive models. Two series of model equations (5+5) were retrieved according to two alternative regression approaches.
► The automatically-generated training subsets (AGTS) was the most effective being able to yield more accurate predictions of fibrate agonist potencies and to automatically filter out entries affected by electronic modulation of the binding properties.
► On the other hand, the most performing model 10 was able to provide for accurate ranking of the fibrates whose binding properties are mainly affected by steric effects. The potential employment of this model in the early screening of newly designed fibrates is currently under investigation in our group.