Strength in numbers: achieving greater accuracy in MHC-I binding prediction by combining the results from multiple prediction tools.

摘要: Background: Peptides derived from endogenous antigens can bind to MHC class I molecules. Those which with high affinity invoke a CD8+ immune response, resulting in the destruction of infected cells. Much work immunoinformatics has involved algorithmic prediction peptide binding various MHC-I alleles. A number tools for have been developed, many are available on web. Results: We hypothesize that peptides predicted by more likely than those just one tool, and likelihood particular being binder is related predict it, as well accuracy tools. To this end, we built tested heuristic-based method making MHC-binding predictions combining results multiple The predictive performance each individual tool first ascertained. These data used derive weights such better given greater credence. combined was evaluated using ten-fold cross-validation found signicantly outperform when specificity threshold used. It performs comparably best-performing at lower thresholds. Finally, it also outperforms combination linear discriminant analysis. Conclusion: facilitates scanning large proteomes potential epitopes, yielding actual high-affinity binders while reporting very few false positives.