Knowledge-based structure prediction of MHC class I bound peptides: a study of 23 complexes

Area of Science:

• Structural Biology
• Immunoinformatics
• Computational Chemistry

Background:

• Peptide binding to MHC molecules is essential for T-cell recognition and immunogenicity.
• Accurate prediction of peptide binding is vital for rational peptide vaccine design.
• Existing sequence-based methods are limited; structural considerations are needed to improve predictive algorithms.

Purpose of the Study:

• To develop a computational algorithm for accurate and fast modeling of peptide structure within the MHC-binding groove.
• To improve the prediction of peptide-MHC binding by incorporating structural information.
• To assess the influence of various parameters on prediction quality.

Main Methods:

• Utilized 23 solved peptide-MHC class I complexes for structural data.
• Developed a modeling algorithm using peptide backbones and MHC structures as templates.
• Employed a rotamer library and the 'dead end elimination' approach for sidechain conformation prediction.
• Used a simple energy function to select favorable rotamer combinations and correct backbone structures.

Main Results:

• The algorithm correctly identified 85% (92%) of all (buried) sidechains and selected correct backbones using a specific rotamer library.
• Cross-validation showed 70% (78%) of all (buried) residues correctly predicted, along with most backbones.
• Peptide sidechain interactions had a negligible effect on prediction quality.

Conclusions:

• Peptide sidechain structure is primarily determined by MHC interactions and peptide backbone, not sidechain-to-sidechain interactions.
• The proposed methodology successfully selected correct peptide backbones from a limited set.
• Performance under cross-validation indicates the rotamer library needs further refinement with more data for improved representation.