An iterative algorithm for converting a class II MHC binding motif into a quantitative predictive model.

Abstract

Biochemists and molecular biologists have suggested motifs for characterizing the binding of peptide fragments and class II major histocompatibility complex (MHC) molecules based on laboratory results and crystal structures. The iterative algorithm presented here converts a suggested motif into a quantitative data-based model. The database accessed consists of peptide fragments known to bind or not bind to class II MHC molecules of particular haplotypes. Stepwise discriminant analysis is utilized to increase or decrease motif coefficients until the resulting motif classifies all binders and non-binders correctly. Stepwise discriminant analysis is a standard multivariate statistical procedure and is available in comprehensive commercial statistical packages. Program 7M of BMDP Statistical Software was used in this study.