In silico prediction of immunogenic T cell epitopes for HLA-DQ8

Abstract

Background: HLA-DQ alleles are involved in the pathogenesis of hypersensitivity reactions and autoimmune disorders, with HLA-DQ8 associated with several human autoimmune disorders. Limited success has been achieved using sequence-based computational techniques for predicting HLA-DQ8-restricted T cell epitopes while accuracy and efficiency of recently developed structure-based models need to be improved. Results: We describe a combined structure-based prediction approach for DQ8-restricted T cell epitope prediction using a recently developed fast and accurate docking protocol, pDOCK, and molecular surface electrostatic potential (MSEP)-based clustering of pMHC binding interfaces. The prediction model was rigorously trained, tested and validated using experimentally verified DQ8 binding and non-binding peptides. High MHC-binding prediction accuracy is validated against independent experimental data (average area under the ROC curve or average A ROC>0.94). Our model also predicts all binding registers correctly and known T cell activators with a positive predictive value (PPV) of 0.91 or 91%. We also studied the binding patterns of DQ8-binding peptides and confirm the existence of peptide epitopes that do not conform to DQ8-specific consensus peptide-binding motifs. Conclusions: We have developed a model that can be successfully applied as a generic protocol for easy in silico identification of HLA-DQ8 binding peptides and thereby potential DQ8-specific T cell epitopes. The current model is therefore applicable for screening potential vaccine candidates from DQ8 binding peptides irrespective of consensus peptide-binding or sequence motifs. We have also illustrated efficient discrimination of different categories of binders from non-binders as well as different categories of pMHC agonists from non-agonists, while accurately predicting the binding registers of DQ8-restricted peptides. This combined approach provides a set of sensitive and specific computational tools to facilitate high-throughput screening of peptides for immunotherapeutic applications such as controlling allergic and autoimmune responses. © 2012 Khan et al. Licensee Nikolai Petrovsky Publishing.