Biophysical Society Thematic Meeting | Hamburg 2022

Biophysics at the Dawn of Exascale Computers

Poster Abstracts

57-POS Board 57 CALCULATING RELATIVE PROTEIN-LIGAND BINDING AFFINITIES WITH THE ACCELERATED WEIGHT HISTOGRAM METHOD - A BENCHMARK STUDY Sebastian Wingbermühle 1 ; Erik Lindahl 1 ; 1 KTH Royal Institute of Technology, Science for Life Laboratory, Stockholm, Sweden To identify a lead compound, hundreds of thousands of drug candidates are synthesized and tested in high-throughput screening assays. The corresponding time and resources could be saved if a computational workflow combining docking and atomistic molecular dynamics (MD) simulations scanned the initial compound library and recommended a set of tens of potential lead compounds to be actually synthesized. To this end, the drug candidates scoring best in the docking step are ranked according to their relative binding free energies estimated in MD simulations. In the context of the EU project LIGATE, the following approach to calculate relative binding free energies is applied here: For each pair of ligands, an alchemical path converting them into each other is defined and parameterized in λ. The Accelerated Weight Histogram method (AWH) frequently updates the value of λ during the MD simulation and adapts a bias potential on the fly such that each value of λ is sampled according to a target distribution. With th e uniform sampling of λ targeted here, the bias potential converges to the free energy profile along λ. This approach is tested on a benchmark set including in total 13 proteins and approximately 500 ligand pairs for which experimental relative binding free energies and previous simulation results (equilibrium free energy perturbation and non- equilibrium thermodynamic integration) are available. As soon as the relative protein-ligand binding affinities obtained with AWH can be proven to be precise and reliable estimates with this benchmark, the AWH simulations run with GROMACS will be combined with docking simulations performed with LiGen to provide a workflow that scales well on modern HPC machines and can handle several thousands of drug candidates.

106