Recent advances in B-cell epitope prediction methods

We review recent advances in computational methods for B-cell epitope prediction, identify some gaps in the current state of the art, and outline some promising directions for improving the reliability of such methods

Yasser EL-Manzalawy; Vasant Honavar


Scholarcy highlights

  • Identification of epitopes that invoke strong responses from B-cells is one of the key steps in designing effective vaccines against pathogens
  • The B-cell conformational epitope predictor proposed by Rapberger et al works as follows: Fast atomic density evaluation is applied to select an antibody among a library of 26 available antibodies showing best shape complementarity to the target antigen; FastContact algorithm is used to identify the most likely interaction site between the selected antibody and the target antigen; Antigen residues that show a decrease in relative solvent accessible surface area of at least 20% in the complex are predicted as belonging to a discontinuous epitope
  • Special care must be exercised to ensure that the datasets used to train and evaluate the predictors are of high quality:
  • ‚óŹ Constructing Non-Redundant datasets: Redundant antibody-antigen complexes should be eliminated from the datasets of B-cell epitopes used to train and evaluate B-cell epitope predictors
  • The uniqueness of epitope sequences is not a sufficient condition for non-redundancy of the dataset because a pair of unique linear B-cell epitopes can share a high degree of pairwise sequence similarity
  • Input sequence windows ranging from 10 to 20 amino acids flanking the target residue, were tested and the best performance, 66% accuracy, was obtained using a window size of 16 amino acids
  • Unless additional similarity reduction steps are taken to eliminate similar epitopes from the dataset, the predictive performance of B-cell epitope predictors estimated using cross-validation on such datasets can be overly optimistic; and in some cases, lead to false conclusions regarding the performance of different predictors relative to each other

Need more features? Save interactive summary cards to your Scholarcy Library.