Interaction between antibodies and antigens in B-cells is crucial for our bodies to neutralize pathogenic molecules. Identifying antigen epitopes, amino acid residues on the antigen is important to learn how they interact, uncover the mechanisms in play, and understand humoral immunity. Knowledge of antibody-antigen interactions is a key step in the process of developing drugs, therapies, and vaccines that utilize this interaction site. In the laboratory, determining the interface residues of an antigen can be difficult, expensive, and time-consuming since it requires the crystallization of the antigen-antibody complex. With the availability of advanced computational methods and public repositories of structures of antigens, we can use computational tools to predict the epitopes of an antigen without needing to crystallize the complex. To achieve the most accurate predictions, we used different computational approaches. There are template-based methods, intrinsic-based methods, and structure-based methods. To overcome the limitations of intrinsic, template, and structure-based methods, we used meta-methods that integrate orthogonal predictors and use method combination methods to acquire results. Our research group developed Meta-DPI that integrates orthogonal approaches with the recently developed docking-based approach DockPred. Meta-DPI combines orthogonal methods by logistic regression, which is more effective for discrete categorical data like a residue’s interfacial score, and random forest. We explore the suitability of Meta-DPI on 275 bound antigens and 175 unbound antigens to accurately predict the epitope of antigens.

Antigen epitope predictions by computational methods

ACS Spring 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.