Integrative Computational Prediction Strategy for Antibody–Antigen Binding: A Case Study on Interleukin-1 Beta

We present a computational strategy to identify critical epitope/paratope residues and predict binding poses in antibody-antigen complexes. The data set consists of 14 different antibody-bound complexes with 9 diverse antigens, including a challenging target Interleukin-1β (IL-1β)─with its four distinct epitope regions─as a rigorous test case. Our strategy includes epitope/paratope site prediction for the antigen and antibody, respectively, followed by molecular docking and molecular dynamics simulations. Central to this computational strategy is essential site scanning analysis (ESSA), a fast and effective elastic network model-based method that determines binding-related essential residues in proteins. Here, ESSA ranks all residues in a given protein by their effect on the intrinsic dynamics of the protein, where the high-scored residues are considered as potential binding sites. In this study, these essential residues were then used to guide docking calculations using ClusPro, significantly improving the accuracy of antibody-antigen pose prediction compared to blind docking. Our molecular dynamics simulation results also show that ESSA-guided docking not only reproduces known binding modes with higher fidelity but also uncovers mechanistically relevant interaction patterns. However, our initial strategy was dependent on knowledge-based epitope guiding; therefore, we further investigated whether ESSA could predict epitope regions in a more automated way and developed a newer workflow, namely, EPIGUIDE, which does not require prior epitope information. EPIGUIDE was applied to 11 diverse antibody-bound complexes, and successful results were obtained. Additionally, ESSA was able to predict at least one residue in the epitope region without any prior information for the 10 cases. We also compared our results with AlphaFold3 predictions and obtained similar success rates. Our method achieves practical throughput by leveraging the efficiency of coarse-grained modeling in ESSA, bypassing the high computational cost of all-atom physics-based simulations for the initial screening. This enables rapid screening of key binding regions, making it suitable for guiding experimental validation.