Validation Tools for Predicted Linear B-Epitopes: Surface Accessibility

Abstract

Identifying B-cell epitopes plays an important role in vaccine design, immunodiagnostic tests, and antibody production. Therefore, computational tools for reliably predicting B-cell epitopes are highly desirable. In this article the possibility of usage of accessible surface scores of peptides as a validation tool is studied. Janin et al. determined empirical amino acid accessible surface probabilities of twenty amino acids. With these fractional surface probabilities for amino acids, a surface probability (S) at sequence position n can be found using a formula given by Emini et. al. When a peptide is uploaded to the Emini Surface Accessibility Prediction in iedb Analysis Resource, prediction tool separates residues into two groups buried, and surface according to the average of S_ns. To create a criterion to decide whether a given peptide is a linear b-epitope or not, for 344,121 b-epitopes downloaded from iedb database, average buried and exposed probabilities, as well as the ratio ρ of averages for these b-epitopes are computed. The same is done for 111,306 artificially created non epitopes. It is seen that for b-epitopes, the ratio ρ is significantly larger than the ratio ρ for the non epitopes. Therefore if ρ is larger, the peptide is more likely is a b-epitope, and this property can be used as to rank peptides while choosing the most probable linear b-epitope from a long list.