Predictive tools for structure-based drug design: A comparative study of force field-based scoring functions

COMP 193

Pablo Englebienne, and Nicolas Moitessier, Department of Chemistry, McGill University, 801 Sherbrooke street West, Montreal, QC H3A 2K6, Canada
Docking methods are widely used in modern drug design and development. The limitations of these techniques include inaccurate scoring of docked poses and the neglect of side-chain flexibility in the protein receptor. As our first step towards a program for the development of predictive tools in drug design and development, we compared the accuracy in the prediction of binding affinity with numerous commonly used force fields with diverse parameterization. The entropic contribution to the free energy of binding is modelled by an estimation of the solvation energy, as well as a penalty for freezing rotatable bonds. The training set included three different protein targets in complexes with ligands showing large structural diversity and a wide activity range. The output of this preliminary research is being applied to the production of a coarse-grained force field, which will be able to account for side-chain flexibility in the protein target.