Pharmaceutical drug discovery is time-consuming and expensive, with each new drug brought to market now costing roughly $1 billion on average. This cost is driven by the difficulty of drug discovery, and in part by the amount of trial and error involved in the process of finding initial "hits" which modulate the function of a biomolecule, and then refining these into "leads" which have adequate affinity for the biomolecular target and other desirable properties. Computational methods ideally could guide this process, reducing the amount of trial and error involved by suggesting hits in advance of experiment and predicting chemical modifications which will improve these into leads, enhancing affinity while maintaining drug-like properties. But current computational methods are not adequate to change the discovery process in this way. Recent innovations in alchemical free energy calculations based on molecular simulations show considerable promise at reaching the level of accuracy needed to help drug discovery, but these simulations require considerable expertise to set up and conduct, and a great deal of computer power. This proposal focuses on lowering these barriers, providing a new approach to automatically plan and set up these calculations, and improved computational efficiency. Alchemical free energy calculations are one of the most physically realistic computational approaches available, and one of the most promising in terms of accuracy. This project's aims are to (1) develop a new tool to automate setup of relative binding free energy calculations for drug lead optimization;(2) efficiently calculate ligand binding mode occupancies, dramatically reducing the computational expense of binding free energy predictions;and (3) use these techniques to guide experimental drug discovery of histone methyltransferase inhibitors, which show considerable promise as potential anti-cancer drugs. While considerable effort has gone into alchemical free energy calculations, one innovative aspect of this work is the focus on predicting binding mode as well as binding affinity. This is handled by using fast docking methods, in combination with exploratory simulations, to identify a variety of stable ligand binding modes, then including all of these in binding free energy calculations, so that bound structures of individual inhibitors need not be known in advance. This work will speed up promising tools for affinity calculation, and improve automation so that they can more easily be applied to problems in drug discovery. The long-term goal of these techniques is to change the early stage drug discovery process by providing robust computational affinity predictions, and this work provides an important step in that direction.
Pharmaceutical drug discovery produces dramatic public health benefits through new and improved treatments for common diseases and disorders, but it is an expensive, time-consuming process involving much trial and error, and failure is common. This work focuses on dramatically improving computer tools to predict interactions between small molecules and molecular machines. The proposed approaches will help guide the development of new drugs, resulting potential public health rewards.