A method to predict quantitatively peptide binding to HLA DRB1*0401 has been developed using a data set of the relative contributions of each of the naturally occurring amino acids in the context of a simplified peptide back-bone. The prediction assumed that the relative role of each of the peptide side chains could be treated independently and could be measured by assaying each of the 20 naturally occurring amino acids at the central 11 positions of a 13-residue peptide previously shown to contain the minimal requirements for high-affinity binding to HLA-DR proteins. The resultant database was shown to have predictive value when tested on a set of 13 unrelated peptides known to bind DRB1*0401 with a wide range of apparent affinity. The database was tested further by analyzing myelin basic protein. All 13 amino acid peptides containing a hydrophobic amino acid at the third position were synthesized and assayed for binding purified DRB1*0401. In every case, the measured affinity correlated with the predictive values within the experimental error of the assays. Finally, the ability to predict peptide binding to MHC class II molecules was shown to help in identifying T cell determinants. The specificity of DRB1*0401-restricted T cell hybridomas against human serum albumin corresponded to two peptides, predicted and shown to bind the class II protein with high affinity.