(686e) Development of a Computational Tool for the Identification of Viable Regions of Proteins to Target for Binding

The last few years have seen the transformation of computational binding protein design. Novel methods like RFdiffusion, AlphaProteo, and BindCraft are designing varied binding proteins with much higher experimental success rates than previous approaches. As the use of these and other tools grows in the coming years, there will be an emerging need for the automated identification of which regions of the bound proteins (i.e., epitopes) are best to target.

In previous work, we have identified Expected Persistent Pairwise Interaction (EPPI) features, which are properties of protein-protein complexes that are expected to persist over time and contribute meaningfully to the binding energy between the proteins. The EPPI features were demonstrated to provide significant benefits at distinguishing real protein complexes from decoy complexes. As the EPPI features are properties of a complex, they cannot be calculated directly for an epitope. Through analysis of antibody-protein complexes, we have identified the features of epitopes that allow for the creation of favorable EPPI properties upon binding by an antibody. This presentation will describe the identification of the important epitope features, our development of a computational tool to automatically scan a candidate antigen for the most promising epitopes, and a computational assessment of the results of using RFdiffusion to target favorable and detrimental epitopes for a varied set of antigens.