Biological Applications of QM/MM Calculations with MM Polarization

David Řeha, Himanshu Sharma and Christopher A Reynolds

Department of Bilogical Sciences, University of Essex, Wivenhoe Park, Colchester, CO4 3SQ, United Kingdom

Hybrid quantum mechanical/molecular mechanics (QM/MM) calculation can be widely used for the study of the biological systems. The effect of polarization on the charge distribution is usually included in the QM part only and MM part is often neglected due to difficulties of the implementation. The MM polarization can however play an important role. Furthermore, there is inconsistence since the one part of the system (QM part) is polarized and the other (MM part) is not. Therefore we have introduced simple approach to treat the polarization of the MM part of the model. The approach is based on the treatment of polarization by induced atomic charges instead of induced dipoles. In this method, the induced dipoles (calculated from atomic polarizabilities and electrostatic potential arising from QM part) are represented by induced charges on the MM atom itself plus its neighboring atomic sites [1]. The main advantage of this approach is the easy implementation to the existing MM programs, since the induced charges are simply added to the permanent MM atomic charges and there is no need to implement the evaluation of charge-dipole interactions. We have implemented the method in various QM/MM programs.

We have applied the method for the study of the enzymatic reaction of chorismate to prephenate within the chorismate mutase. The polarization has stabilizing effect on the transition state and we have observed the decrease of the activation energy by 5-7 kcal/mol.

We have also applied the MM polarized QM/MM method in the attempt to improve the results of the ligand docking in the case of the fragment based drug design. Since the smaller fragments bind more weakly to the protein than the bigger drug like molecules, the more accurate docking is required. The standard docking by program Glide was performed on the series of the fragments with known crystal structure [2]. The docking results were then significantly improved by rescoring of the poses using QM/MM calculations and by re-docking the ligands using the polarized charges of both the ligand and the protein. We have successfully predicted 11 out of 12 structures when using our approach, compare to just 5 out of 12 structures when the standard docking by Glide was used.

1. Ferenczy, G. G. Reynolds, C. A. J. Phys. Chem. A , 105, (2001), 11470.

2. Congreve et al., J. Med. Chem, 51, (2008), 3661.