Computational epitope mapping of class I fusion proteins using low complexity supervised learning methods

Antibody epitope mapping of viral proteins plays a vital role in understanding immune system mechanisms of protection. In the case of class I viral fusion proteins, recent advances in cryo-electron microscopy and protein stabilization techniques have highlighted the importance of cryptic or ‘alternative’ conformations that expose epitopes targeted by potent neutralizing antibodies. Thorough epitope mapping of such metastable conformations is difficult but is critical for understanding sites of vulnerability in class I fusion proteins that occur as transient conformational states during viral attachment and fusion. We introduce a novel method Accelerated class I fusion protein Epitope Mapping (AxIEM) that accounts for fusion protein flexibility to improve out-of-sample prediction of discontinuous antibody epitopes. Harnessing data from previous experimental epitope mapping efforts of several class I fusion proteins, we demonstrate that accuracy of epitope prediction depends on residue environment and allows for the prediction of conformation-dependent antibody target residues. We also show that AxIEM can identify common epitopes and provide structural insights for the development and rational design of vaccines.Author summary: Efficient determination of neutralizing epitopes of viral fusion proteins is paramount in the development of antibody-based therapeutics against rapidly evolving or undercharacterized viral pathogens. Advances in the determination of viral fusion proteins in multiple conformations with ‘cryptic epitopes’ during attachment and fusion has highlighted the importance of epitope accessibility due to viral fusion protein flexibility, a physical trait not accounted for in previous B-cell epitope prediction methods. To consider how protein flexibility might influence antigenicity, viral fusion proteins must have been determined in conformations that correspond with multiple stages of attachment and/or fusion–and have been extensively subjected to B-cell epitope mapping techniques. Despite advances in cryptic epitope determination, the available data is limited to a subset of class I fusion proteins that meet the above criteria. This poses a challenge to computational epitope mapping in generating an informative model that avoids overfitting. Here, we discuss a limited set of descriptive features, that when used in a variety of low complexity classifier models, matches or outperforms other publicly available B-cell epitope prediction methods in out-of-sample tests. From the models we tested, we use the linear regression model to highlight structural insights of epitopes and to demonstrate how this model may provide a novel approach to assess structural changes of antigenicity between viral fusion protein homologues.