Challenges in Protein Structure Prediction and Drug Discovery

Alignment or superposition of multiple flexible ligands in 3D is a key step in rational ligand-based drug design, pharmacophore elucidation and 3D QSAR analysis. We have recently introduced Atomic Property Fields methodology, which utilizes continuous Gaussian-based multicomponent potentials to represent the distributions of physico-chemical atomic properties. Calculation of APF pseudo-energy provides a sensitive measure of similarity between these distributions for two or more molecules in 3D. Because the potentials are continuous and smooth, gradient minimization can be used to achieve locally optimal fit, while global sampling is performed using Monte-Carlo to identify globally optimal superposition. The method is shown to successfully identify non-trivial superpositions, and to be effective in VLS and 3D QSAR tests. The approach is next extended to the superposition and comparison of the binding sites in protein receptors. Atom set forming a shell around the binding site is treated as a 'molecule' to be superimposed. Application of the APF superposition approach to distantly related active sites shows the ability of the method to identify and superimpose common features despite wide overall divergence of the receptor structures. Furthermore, APF pseudo-energy can serve as a measure of similarity between pairs of binding sites. Sites similar to a query are successfully identified in a database of binding sites. Finally, by converting the APF score into a distance measure and applying hierarchical clustering, all binding sites in a database can be clustered into families which correspond well to the established functional classes.