Analysis and prediction of protective continuous B-cell epitopes on pathogen proteins

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Peptide-based diagnostics and therapeutics mimicking protein antigens are gaining interest.
• Current peptide selection relies on T-cell epitope prediction, structural data, or B-cell epitope predictions, often insufficient without structural information.
• Rational selection of peptides for antibody production is challenging, especially when many antigens are available, necessitating efficient methods to reduce synthesis efforts.

Purpose of the Study:

• To develop and validate a computational method for selecting and ranking peptides for antibody stimulation.
• To improve the rational selection of peptides by integrating B-cell epitope prediction with antigen variability and post-translational modification (PTM) pattern conservation.
• To investigate the application of machine learning (ML) for enhanced epitope prediction.

Main Methods:

• Integration of B-cell epitope prediction with antigen variability and PTM prediction.
• Development of a cumulative parameter approach to identify functionally relevant protein regions.
• Application of machine learning methods, including Random Forests, for improved predictive power.
• Creation of a curated dataset for validating protectivity classification.

Main Results:

• Protein regions yielding functionally altering antibodies show increased cumulative parameter sums.
• Maximizing antigenicity, minimizing variability, and minimizing PTM likelihood identified relevant sites.
• The approach achieved Area Under the ROC Curve (AROC) values of at least 0.65 for approximately 50% of antigens.
• Machine learning improved classification accuracy, achieving up to 83% separation in an equilibrated dataset.
• Results suggest at least two distinct classes of bioinformatically detectable epitopes requiring separate prediction.

Conclusions:

• A computational method was developed for automatic selection and ranking of peptides to stimulate protective or functionally altering antibodies.
• Integrating variability, PTM conservation, and B-cell antigenicity enhances rational peptide selection over random methods.
• The method performs substantially better for certain antigens, optimizing for invariable and unmodified peptides beneficial for diagnostics and therapeutics.
• Machine learning, particularly Random Forests, shows potential for further improving predictive capabilities.