Computational antigenic epitope prediction by calculating electrostatic desolvation penalties of protein surfaces

The prediction of antigenic epitopes on the surface of proteins is of great importance for vaccine development and to specifically design recombinant antibodies. Computational methods based on the three-dimensional structure of the protein allow for the detection of noncontinuous epitopes in contrast to methods based on the primary amino-acid sequence only. A method recently developed to predict protein-protein binding sites is presented, and the application to predict putative antigenic epitopes is described in detail. The prediction approach is based on the local perturbation of the electrostatic field at the surface of a protein due to a neutral probe of low dielectric constant that represents an approaching binding partner. The calculated change in electrostatic energy corresponds to an energy penalty of desolvating a protein surface region, and antigenic epitope surface regions tend to be associated with a lower penalty compared to the average protein surface. The protocol to perform the calculations is described and illustrated on an example antigen, the outer surface protein A of Borrelia burgdorferi, a pathogenic organism causing lyme disease.