The physical basis of computer-aided drug design: assessing and advancing the accuracy of binding affinity calculations*

Predicting the binding thermodynamics of proteins and small organic molecules is a key aim in computer-aided drug design. I will provide an introduction to this problem in physical chemistry, the statistical mechanical framework for models of binding, and computational approaches based on this framework. I will then describe some interesting and unexpected outcomes of blinded prediction challenges in this field and what they indicate about steps required to improve accuracy, notably the need for improve potential functions, also known as force fields. Finally, I'll present the approaches we are taking, in coordination with the Open Force Field Initiative, to improve the accuracy of force fields and thus speed the discovery of new medications.