Customized versus universal scoring functions: application to class I MHC-peptide binding free energy predictions

Area of Science:

• Computational chemistry
• Molecular modeling
• Immunology

Background:

• Predicting peptide-MHC binding is crucial for understanding immune responses.
• Class I human major histocompatibility protein (HLA) molecules present peptides to cytotoxic T lymphocytes.
• HLA-B*2705 is associated with autoimmune diseases, making its peptide binding properties of significant interest.

Purpose of the Study:

• To evaluate the performance of a tailor-made free energy scoring method (Fresno) against established universal scoring functions.
• To assess the accuracy of different scoring functions in predicting the binding affinity of peptides to HLA-B*2705.

Main Methods:

• A dataset of 26 peptides with known binding affinities to HLA-B*2705 was used.
• The Fresno scoring method was compared against six universal scoring functions: Chemscore, Dock, FlexX, Gold, Pmf, and Score.
• Binding affinity predictions were generated using each scoring function.

Main Results:

• The Fresno scoring method significantly outperformed all six universal scoring functions in predicting peptide binding affinity.
• Fresno demonstrated higher accuracy and reliability in assessing the interactions between peptides and HLA-B*2705.

Conclusions:

• The tailor-made Fresno scoring method offers a more accurate approach for predicting peptide binding to HLA-B*2705 compared to general scoring functions.
• Fresno's superior performance suggests its potential utility in immunological studies and drug discovery targeting HLA-B*2705.