SEPPA 3.0-enhanced

SEPPA 3.0-enhanced predicts spatial B-cell epitopes on protein antigens, explicitly modeling N-linked glycosylation effects to improve epitope identification for immunotherapy and vaccine design.

Key Features:

• Enhanced glycoprotein antigen prediction: Tailored to handle N-linked glycoproteins and account for the impact of glycosylation on B-cell epitope prediction.
• Updated parameters and classifiers: Incorporates an updated dataset with micro-environmental features including glycosylation triangles and glycosylation-related amino acid indexes as classifiers.
• Logistic regression model: Retains the logistic regression framework from SEPPA 2.0 for antigenicity prediction.
• Neighboring antigenicity calibration: The logistic regression framework is calibrated based on neighboring antigenicity information.
• Validation and performance metrics: Achieved AUC 0.794 by 10-fold cross-validation on an internal dataset; for general protein antigens AUC 0.740 and BA 0.657; for independent glycoprotein antigens AUC 0.749 and BA 0.665, outperforming peers in the glycoprotein domain.

Scientific Applications:

• Immunotherapy development: Predicts B-cell epitopes on glycoproteins to aid identification of antigenic regions relevant to antibody-based therapies.
• Vaccine design: Informs selection of epitope targets on glycoproteins where glycosylation influences antigenicity.
• Antigenicity studies of post-translational modifications: Enables analysis of how glycosylation-related micro-environmental features affect B-cell epitope mapping on complex antigens.

Methodology:

Integrates an updated dataset with micro-environmental glycosylation features (including glycosylation triangles and glycosylation-related amino acid indexes) and applies a logistic regression model calibrated on neighboring antigenicity, with performance evaluated by 10-fold cross-validation and independent testing reporting AUC and BA metrics.