Relative binding free energies for drug discovery: applications to p38 MAP kinase

Abstract

In the early stages of drug development, it is crucial to gain an accurate understanding of the binding affinity of a lead compound to the target of interest. Over the years, computer aided drug design has proven useful through the ability to perform simulations yielding the binding free energy of drug-like molecules to proteins. A popular approach is to use classical molecular dynamics (MD) methods to perform alchemical relative binding free energy calculations (RBFEs), despite being aware that these conventional methods are prone to poor sampling and force field errors. Overcoming these challenges is of particular relevance for systems where the protein has a high degree of flexibility as well as when ligands have multiple rotatable bonds. One such system is p38 MAP Kinase, which has flexible loops close to the binding site and several rotatable torsions in the ligand structure. The work presented in this thesis aims to investigate the sampling and force field quality in relative binding free energy calculations with p38 in an effort to understand how complexities in such a system may be addressed in the context of drug discovery. In Chapter 3, force field torsion parameters are reparameterised for a ligand dataset using a pipeline based on fitting to 2D energy surfaces derived from Density Functional Theory (DFT). This is in an effort to improve the ability of the force field to capture important dynamics that may affect free energy estimates. The performance of the original force field in free energy (FEP) calculations alongside two variations incorporating the new torsional parameters are presented. In Chapter 4, a non-equilibrium free energy method named Non-Equilibrium Switching (NES) is applied to the p38 dataset. We show that the correlations between predicted and experimental RBFEs are influenced by the dynamic range of free energies derived from the chosen perturbation edge map. Additionally, we show the bias owing to the initial crystal structure in the calculated free energies on a per edge basis by running NES using nine different initial crystal structures. In Chapter 5, we explore a novel method named Fully Adaptive Simulated Tempering (FAST), which combines Sequential Monte-Carlo (SMC) and Irreversible Simulated Tempering (IST). This method enables user specified torsional enhanced sampling over long timescales with an adaptive

More information

Identifiers

ORCID for Jonathan Essex: orcid.org/0000-0003-2639-2746

ORCID for Jeremy Frey: orcid.org/0000-0003-0842-4302

Catalogue record

Export record