Free energy calculations in drug discovery

Protein-ligand binding free energy is a crucial property used in the discovery and optimisation of novel drugs. In the context of computer-aided drug design, being able to calculate accurate, precise and reproducible binding free energies in silico is essential in terms of reducing the amount of trial-and-error synthetic effort required to develop a candidate drug. More potent and selective compounds will bind strongly to the desired target, and show weak binding to related proteins. Recent commercial software products have pushed the use of binding free energy calculations into mainstream development pipelines within the pharmaceutical industry. Even with pre-screening of suitable target systems, however, the required predictive accuracy of 1 kcal mol-1 is only attained in around half of cases. Determining whether sufficient accuracy has been achieved for a particular system is currently not possible a priori and requires expensive and time-consuming experimental validation.

In this project the accuracy of these free energy calculations in the context of the pharmaceutical industry will be explored. We will develop a large dataset of test systems based on published data and determine the sensitivity of the calculated free energies to the choice of simulation software, parameters, free energy calculation method and system setup. Based on this sensitivity analysis, the project objectives are twofold:
1. To identify the primary causes of inaccuracy in protein-ligand binding free energy calculations. Through this we will be able to identify a priori on which systems these approaches will be successful. This information will be extremely valuable, allowing those cases where the methods will be useful to be identified without requiring extensive and expensive experimental validation.
2. To improve free energy calculation methodology to address these sensitivities.