PPARδ is a ligand-activated receptor that dimerizes with another nuclear receptor of the retinoic acid receptor family. The dimers interact with other co-activator proteins and form active complexes that bind to PPAR response elements and promote transcription of genes involved in lipid metabolism. It appears that various natural fatty acids and their metabolites serve as endogenous activators of PPARδ; however, there is no consensus in the literature on the nature of the prime activators of the receptor. In vitro and cell-based assays of PPARδ activation by fatty acids and their derivatives often produce conflicting results. The search for synthetic and selective PPARδ agonists, which may be pharmacologically useful, is intense. Current rational modelling used to obtain such compounds relies mostly on crystal structures of synthetic PPARδ ligands with the recombinant ligand binding domain (LBD) of the receptor. Here, we introduce an original computational prediction model for ligand binding to PPARδ LBD. The model was built based on EC50 data of 16 ligands with available crystal structures and validated by calculating binding probabilities of 82 different natural and synthetic compounds from the literature. These compounds were independently tested in cell-free and cell-based assays for their capacity to bind or activate PPARδ, leading to prediction accuracy of between 70% and 93% (depending on ligand type). This new computational tool could therefore be used in the search for natural and synthetic agonists of the receptor.
© 2014 The British Pharmacological Society.