BPS2026 – A physics-based computational approach for absolute binding affinity calculations: A hemagglutinin-65C6 antibody application

The dynamic nature of protein structures and protein-drug interactions is a key reason why many commercially available tools fail to accurately predict drug binding affinities. Accurate, physics-based approaches can accelerate drug discovery by filtering out weak binders early, thereby reducing both time and cost. To address this challenge, we present a novel physics-based method for estimating binding affinities that is both accurate and computationally efficient. The novelty of our algorithms lies in (1) the application of a rigorous theoretical framework for binding free energy estimation, specifically the grid potential of mean force, (2) the use of non-parametric reweighting to address orthogonal degrees of freedom, (3) the development of highly scalable, loosely coupled multiple-copy algorithms, and (4) the application of orientation quaternion formalism to describe the relative orientation of ligands with respect to proteins. As a case study, we applied this framework to antibody-antigen interactions, focusing on the 65C6 antibody, which neutralizes a wide range of highly pathogenic H5N1 influenza strains through binding to the hemagglutinin (HA) protein. Our simulations of the 65C6-HA complex yield binding affinity estimates that provide a foundation for rational antibody design against H5N1 and establish a generalizable strategy for accurate antibody-antigen binding predictions.