Identification of ligand-binding pockets in proteins using residue preference methods.

Identification of ligand-binding pockets in proteins is pivotal to protein function definition and drug discovery. In this study, we focus on determining the binding pockets in proteins for potential ligands without any a priori knowledge. Three methods based upon residue preference concept are proposed to predict ligand-binding pockets, where we deal with three types of residue preference (residue based, atom based and atom-contact-pair based preference), respectively. Two test sets were chosen to examine the proposed methods. Two different identification rules (named Top1 and Top2) are used to detect ligand-binding pockets. The results show that the atom-contact-pair method has good accuracy and high efficiency, better than the other two methods. By means of preference analysis for amino acids and atom-contact-pairs, we find that Gly and atom-contact-pairs on aromatic residues appear at ligand-binding pockets more frequently. The former favors pocket flexibility, and the latter shows that aggregate hydrophobic surface may play an important role in complex formation.