The identification of ligand binding sites in a protein is the key to several applications in the bioinformatics area, such as molecular docking, de novo drug design and structural identification of functional sites. In recent years it has been demonstrated that the concept of frustration is related to several functional aspects of proteins. For this reason, we present a new method that predicts protein-ligand interaction sites based on frustration.
A non-redundant data set of all monomeric enzymes with protein-ligand binding sites annotations were downloaded from the BioLiP database (n = 1007) to characterize protein-ligand interaction sites. Protein structures were downloaded from the Protein Data Bank (PDB) and frustration patterns were calculated using the Frustratometer tool. Characterization of the energetic patterns of all the selected proteins was performed. In addition, it was calculated the percentage of highly frustrated contacts and the solvent accessible surface (SASA) of the residues that are implicated in protein-ligand interaction.
We found that protein residues that directly interact with ligands are enriched in highly frustrated interactions. The method of FrustraPocket was used to predicted protein–ligand binding sites using frustration and SASA as parameters. In order to compare our method, the fpocket tool was used. It was found that Frustrapocket had a higher success rate than fpocket. FrustraPocket is still under development and it will be available soon for the scientific community.