A simple method to predict protein-binding from aligned sequences--application to MHC superfamily and...

Area of Science:

• Immunology
• Structural Biology
• Bioinformatics

Background:

• The MHC superfamily (MhcSF) includes immune MHC class I (MHC-I) and related proteins with diverse functions.
• beta2-Microglobulin (B2M) binding is crucial for MHC-I surface expression and function, but its interaction with MHC-I-like proteins varies.
• Predicting B2M binding is key to deciphering MhcSF protein functions and identifying disease-related mutations.

Purpose of the Study:

• To develop a predictive tool for B2M binding in MhcSF proteins.
• To leverage IMGT unique numbering for precise feature selection and classification.
• To understand molecular recognition mechanisms and identify functionally significant mutations within the MhcSF.

Main Methods:

• A simple-Bayes classifier combined with IMGT unique numbering was employed.
• Discriminant binary features were selected by associating alignment positions with amino acid groups.
• Classifier performance was assessed using leave-one-out procedures for novel proteins, species, and receptor types.

Main Results:

• The dataset comprised 806 allelic forms of 47 MhcSF proteins across 9 receptor types and 4 mammalian species.
• Eighteen discriminant features, including B2M contact sites and stabilizing elements, were identified.
• High prediction accuracies were achieved: 98% for new proteins, 94% for new species, and 70% for new receptor types.

Conclusions:

• The developed classifier accurately predicts B2M binding across diverse MhcSF proteins.
• The tool's application to lower vertebrate MHC-I proteins suggests conserved B2M binding and surface expression mechanisms.
• This approach demonstrates broad applicability for predicting protein functions and interactions within protein superfamilies.