Improving the accuracy of MM-GBSA binding energy predictions through parameter optimization

COMP 40

B. Woody Sherman, sherman@schrodinger.com, Jianxin Duan, duan@schrodinger.com, and Matthew P. Repasky, repasky@schrodinger.com. Schrödinger, Inc, 120 West 45th Street, New York, NY 10036-4041
Recent work has shown that the MM-GBSA method is capable of rapidly and accurately predicting trends in binding activity of molecules within a congeneric series. While these findings are encouraging, little work has gone into optimizing parameters within the MM-GBSA methodology to improve the reliability and accuracy of the calculations. In this work, we study factors influencing the predictive capabilities of MM-GBSA, such as the internal molecular dielectric constant, the extent of allowed protein flexibility, and the ligand charge model. Diverse targets and ligand sets are studied in order to better understand the optimal settings to be used in MM-GBSA studies.