GradDock: rapid simulation and tailored ranking functions for peptide-MHC Class I docking.

Area of Science:

• Computational biology
• Structural immunology
• Bioinformatics

Background:

• T-cell epitope identification is vital for transplantation, diagnostics, vaccine development, and immunotherapy.
• Structural modeling of peptide-MHC interactions is key to understanding immunological mechanisms.
• Data-driven prediction methods show success, but structural insights remain crucial.

Purpose of the Study:

• To develop GradDock, a rapid and accurate structure-based method for modeling peptide binding to MHC Class I (pMHC-I).
• To enhance the understanding of molecular mechanisms underlying T-cell epitope recognition.

Main Methods:

• GradDock explicitly models unbound peptides and inserts them into the MHC-I groove using steered gradient descent with topological correction.
• Revised Rosetta score terms and developed a novel ranking function specifically for pMHC-I.
• Employed linear programming to optimize score term weights, refining dihedral angles and repulsion for improved modeling quality.

Main Results:

• GradDock achieves five-times faster docking than Rosetta-based methods for pMHC-I.
• The re-weighted Rosetta ranking function in GradDock demonstrates consistently superior accuracy (approx. three-times better on cross-docking sets) compared to standard Rosetta scores.
• The method yields diverse structural conformations, including native-like peptides.

Conclusions:

• GradDock offers a significant advancement in the speed and accuracy of pMHC-I modeling.
• This method facilitates deeper understanding and improved prediction of T-cell epitopes.
• The developed tool and ranking weights are freely available for academic research.