Support vector machine-based prediction of MHC-binding peptides.

Abstract

The use of major histocompatibility complex (MHC) class I binding peptides for immunotherapeutic purposes has shown promising results in recent years. The identification of such peptides mostly starts with predicting MHC-binding peptides, given a protein of interest. An accurate prediction method can reduce the number of peptides that needs to be tested experimentally. This protocol describes in this describes how support vector machines (SVMs) can be used for predicting MHC class I binding peptides. Focus is given on data representation, the concept of cross-validation, and how optimal SVM-specific parameters are obtained.