Novel type of virtual ligand screening on the basis of quantum-chemical calculations for protein-ligand complexes and extended clustering techniques

We propose a novel method for virtual ligand screening to explore drug candidates binding to target proteins. Employing both information on the ligand-residue interactions calculated by the fragment molecular orbital method and the molecular properties represented by the MDL MACCS keys, a couple of advanced clustering analyses on the basis of the self-organizing map and the multi-dimensional scaling are carried out. In comparison to earlier, similar approaches, the present method can provide higher-dimensional, wider viewpoints to look for better inhibitors and to improve them with reducing the possibilities of false-positives and false-negatives for hit or lead compounds, thus accelerating drug designs. The feasibility and usefulness of the proposed methodology are then demonstrated through the application to the complex systems of estrogen receptor and its ligand molecules, in which a molecular modification to improve the binding property of drug candidates is also suggested.