A Relevance Vector Machine Based Quantitative Prediction Method for Mouse Class I MHC Peptide Binding Affinity

The accurate identification of T-cell epitopes remains a principal goal of bioinformatics within immunology. As the immunogenicity of peptide epitopes is dependent on their binding to major histocompatibility complex (MHC) molecules, the prediction of binding affinity is a prerequisite to the reliable prediction of epitopes. Existing computational prediction approaches can be roughly classified into two distinct types, one categorizes peptides as "strong binders" and "weak binders", namely qualitative predictions. And the other makes predictions about the precise binding affinities, namely, quantitative prediction. Recently, the modeling of MHC-peptide binding has come to emphasize quantitative predictions. We proposed a relevance vector machine (RVM) based approach, named RVMMHC for quantitative modeling. Our approach adopted the same encoding scheme as used in SVRMHC but chose relevance vector machine to build the predicting models. When applied to the MHC-peptide binding systems of three mouse class I MHC alleles, the RVMMHC models produced more accurate predictions than those produced by linear models which had represented a significant advance in existing methods such as SYFPEITHI, BIMAS and RANKPER. Further analyses showed that SVRMHC gained slightly better performance than RVMMHC, while RVMMHC had better sparsity and some other advantages. Finally, comparison based on the receiver operating characteristic analyses (ROC) indicated that RVMMHC and SVRMHC out-performed above-outlined methods including additive method in identifying strong binder peptides.