Accurate prediction of binding free-energies remains a central challenge in antibody design, where large, flexible interfaces and subtle energetic differences determine specificity. Conventional tools often fail to capture this complexity, producing unreliable affinity estimates. To address this gap, we have developed a physics-based methodology for absolute binding free-energy (ABFE) estimation that is both rigorous and computationally efficient. Our approach builds on a grid potential of mean force (grid PMF) framework, applies non-parametric reweighting to capture orthogonal degrees of freedom, employs highly scalable multiple-copy molecular dynamics (MD) simulations, and uses an orientation-quaternion formalism to describe molecular orientations in a robust manner. The use of loosely coupled, multiple-copy MD simulations also allows the method to scale efficiently on modern high-performance computing platforms, making it feasible to apply ABFE calculations to large biomolecular systems such as antibody-antigen complexes. Here, we extend our grid PMF methodology to antibody-antigen recognition. Specifically, we applied our protocol to the broadly neutralizing antibody 65C6, which binds hemagglutinin (HA) from highly pathogenic avian influenza H5N1 strains. Our simulations of the HA-65C6 interaction yield absolute binding free-energy estimates that provide a quantitative foundation for rational antibody engineering against influenza. This case study demonstrates the broader potential of our framework. By successfully capturing the dynamic and flexible features of antibody-antigen binding, our approach offers a path toward designing antibodies with tailored affinity and cross-reactivity. Beyond influenza, the method is applicable to antibody development for emerging infectious diseases, immunodiagnostics, and therapeutic interventions, where reliable affinity predictions are urgently needed.
Sauve et al. (2026) studied this question.