Toward Guided Mutagenesis: Gaussian Process Regression Predicts MHC Class II Antigen Mutant Binding

Area of Science:

• Immunology
• Computational Biology
• Biophysics

Background:

• Antigen-specific immunotherapies (ASI) rely on effective antigen peptide presentation by major histocompatibility complex (MHC) molecules.
• Modifying antigen binding to MHC is a key strategy in ASI design, but exploring all mutations is resource-intensive.
• Accurate prediction of peptide-MHC class II (pMHCII) binding affinities for mutations is essential for efficient ASI development.

Purpose of the Study:

• To determine the minimum prior data needed for accurate prediction of pMHCII mutant binding affinities.
• To investigate the efficacy of Gaussian process (GP) regression for predicting relative binding affinities of point mutations.
• To reduce the experimental and computational costs associated with exploring antigen mutations for ASI.

Main Methods:

• Utilized Gaussian process (GP) regression to interpolate pMHCII mutant binding affinities based on residue volume and hydrophobicity.
• Applied GP regression to experimental data from the Immune Epitope Database and theoretical data from NetMHCIIpan and Free Energy Perturbation.
• Evaluated prediction accuracy using R-squared, average error (kcal/mol), and receiver operating characteristic (ROC) AUC for binary classification.

Main Results:

• GP regression predicted binding affinities for nine neutral residues from a six-residue subset with R-squared of 0.62 ± 0.04 and ROC AUC of 0.92.
• Prediction accuracy improved to R-squared of 0.69 ± 0.04 and ROC AUC of 0.94 when predicting seven neutral residues from an eight-residue subset.
• Prediction accuracy was highest for neutral residues at anchor sites without register shift.

Conclusions:

• Gaussian process regression can accurately predict pMHCII binding affinities using limited prior data, significantly reducing costs for ASI design.
• The developed method offers a valuable tool for accelerating the design and optimization of antigen-specific immunotherapies.
• This approach aids in understanding the impact of specific residue mutations on peptide-MHC binding interactions.