A critical cross-validation of high throughput structural binding prediction methods for pMHC

Area of Science:

• Immunology
• Computational Biology
• Structural Biology

Background:

• T-cells recognize antigens via T-cell receptors (TCRs).
• Major Histocompatibility Complex (MHC) presents peptide antigens to TCRs on the cell surface.
• Existing in silico prediction methods primarily rely on amino acid sequences.

Purpose of the Study:

• To develop and evaluate a structural computational approach for predicting peptide-MHC binding.
• To provide insights into the spatial binding geometry of T-cell epitopes.
• To achieve high data throughput with accurate classification of potential binders and non-binders.

Main Methods:

• Utilized side chain substitution (threading) for peptide modeling.
• Employed energy minimization to refine peptide-MHC complex structures.
• Applied scoring functions to evaluate protein/peptide interfaces.
• Performed approximately 83,000 binding affinity prediction runs.

Main Results:

• The structural approach offers insights into spatial binding geometry, unlike sequence-based methods.
• Prediction accuracy for classifying peptides as binders or non-binders ranged up to 75%, depending on the combination of tools used.
• The accuracy is highly dependent on the effectiveness of threading, energy minimization, and scoring steps.

Conclusions:

• The developed structural approach provides a viable method for classifying T-cell epitopes based on their binding potential to MHC molecules.
• This method can guide experimental efforts by identifying potential binders and non-binders.
• The accuracy of prediction is contingent upon the integrated performance of the computational tools employed.