Class II MHC quantitative binding motifs derived from a large molecular database with a versatile iterative stepwise discriminant analysis meta-algorithm

Motivation: The identification of T-cell epitopes can be crucial for vaccine development. An epitope is a peptide segment that binds to both a T-cell receptor and a major histocompatibility complex (MHC) molecule. Predicting which peptide segments bind MHC molecules is the fir st step in epitope prediction. Results: An iterative stepwise discriminant analysis meta-algorithm explores a large molecular database to del-ive quantitative motifs for peptide binding. The applications presented here demonstrate the algorithm's versatility by producing four closely related models for HLA-DR1. Two models use an expert initial estimate and two do not; two models use amino acid residues as the only predictors and two use amino acid groupings as additional predictors. Each model correctly classifies >90% of the peptides in the database.