Enhanced antibody-antigen structure predictionfrom molecular docking using AlphaFold2

Abstract

Predicting the structure of antibody-antigen complexes has tremendous value in biomedical research but unfortunately suffers from a poor performance in real-life applications. AlphaFold2 (AF2) has provided renewed hope for improvements in the field of protein-protein docking but has shown limited success for the medically relevant class of antibody-antigen complexes due to the lack of co-evolutionary constraints. Some research groups have demonstrated the usefulness of the AF2 confidence metrics for assessing the plausibility of protein folding models. In this study, we used physics-based protein docking methods for building decoy sets consisting of low-energy docking solutions that were either geometrically close to the native structure (positives) or not (negatives). The docking models were then fed into AF2 to assess their confidence with a novel composite score based on the pLDDT and pTMscore metrics. We show benefits of the AF2 composite score for rescoring docking poses in two scenarios: (1) a more trivial experiment based on the bound conformations of the antibody and antigen backbone structures, and (2) a more realistic test employing the unbound backbone conformations of the binding partners. Docking success rates improved after AF2 rescoring with particular emphasis on early enrichment of positives at the very top of the re-ranked list of decoys. The AF2 rescoring markedly improved classification of positives and negatives in most systems. Docking models of at least medium quality present in the decoy set, but not necessarily highly ranked by docking methods, benefitted most from AF2 rescoring by experiencing large advances towards the top of the reranked list of models. These improvements, obtained without any calibration or novel methodologies, led to a notable level of performance in antibody-antigen unbound docking that was never achieved previously.