OETMAP: a new feature encoding scheme for MHC class I binding prediction

Abstract

Deciphering the understanding of T cell epitopes is critical for vaccine development. As recognition of specific peptides bound to Major histocompatibility complex (MHC) class I molecules, cytotoxic T cells are activated. This is the major step to initiate of immune system response. Knowledge of the MHC specificity will enlighten the way of diagnosis, treatment of pathogens as well as peptide vaccine development. So far, a number of methods have been developed to predict MHC/peptide binding. In this article, a novel feature amino acid encoding scheme is proposed to predict MHC/peptide complexes. In the proposed method, we have combined orthonormal encoding (OE) and Taylor's Venn-diagram, and have used Linear support vector machines as the classifier in the tests. We also have compared our method to current feature encoding scheme techniques. The tests have been carried out on comparatively large Human leukocyte antigen (HLA)-A and HLA-B allele peptide three binding datasets extracted from the Immune epitope database and analysis resource. On three datasets experimented, the IC50 cutoff a criteria is used to select the binders and non-binders peptides. Experimental results show that our amino acid encoding scheme leads to better classification performance than other amino acid encoding schemes on a standalone classifier.