Rapid, precise and reproducible binding affinity prediction : applications in drug discovery

• Main Author: Jovanovic, Srdan
• Other Authors: Coveney, P.
• Published: University College London (University of London) 2018
• Subjects: 540
• Online Access: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.756212

As we move towards an era of personalised medicine, the identification of lead compounds requires years of research and considerable financial backing, in the development of targeted therapies for cancer. We use molecular modelling and simulation to screen a library of active compounds, and understand the ligand-protein interaction at the molecular level in appropriate protein targets, in a bid to identify the most active lead drug candidates. In recent times, good progress has been made in accurately predicting binding affinities for drug candidates. Advances in high-performance computation (HPC), mean it is now possible to run a larger number of calculations in parallel, paving the way for multiple replica simulations from which binding affinities are obtained. This, then, allows for a tighter control of errors and in turn, a higher confidence in the binding affinity predictions. Here, we present ESMACS (Enhanced Sampling of Molecular dynamics with Approximation of Continuum Solvent) and TIES (Thermodynamic Integration with Enhanced Sampling); a new framework from which binding affinities are calculated. ESMACS performs 25 replica simulations of the same ligand-receptor system with the only difference being the initial momentum of each atom. From this ensemble of trajectories, an extended MMPBSA (Molecular Mechanics Poisson-Boltzmann Surface Area) free energy method is employed. The TIES protocol constitutes 5 replicas simulations per lambda state followed by the integration of the potential derivatives of each lambda state, generating a relative binding affinity. This is all tied together using the BAC (Binding Affinity Calculator) which automates the ESMACS and TIES workflow. ESMACS and TIES, given suitable access to HPC resources, can compute binding affinities in a matter of hours on a supercomputer; the size of such machines therefore means that we can reach the industrial scale of demand necessary to impact drug discovery programmes.