SEPPA-mAb

SEPPA-mAb predicts spatial epitopes on protein antigens to support monoclonal antibody (mAb) design and evaluation.

Key Features:

• High accuracy and low false positive rate: Achieves an accuracy of 0.873 with a false positive rate (FPR) of 0.097 when classifying epitope versus non-epitope residues.
• Fingerprints-based patch model: Implements a fingerprints-based patch model that evaluates structural and physicochemical complementarity between antigen patches and mAb complementarity-determining regions (CDRs).
• Training data: The predictive model is trained on 860 representative antigen-antibody complexes.
• Performance benchmarks: Reported balanced accuracy is 0.635, with comparative AUCs cited for docking-based methods (AUC 0.691) and other epitope prediction tools (AUC 0.730) in independent testing.
• HIV glycoprotein evaluation: In a focused study on 36 independent HIV glycoproteins, reported accuracy was 0.918 with an FPR of 0.058.
• Robustness: Demonstrates robustness when applied to new antigens and modelled antibodies.
• Algorithm lineage: Extends methodological capabilities from SEPPA 3.0.

Scientific Applications:

• Monoclonal antibody design: Identifies spatial epitope positions to guide mAb specificity and efficacy optimization.
• Epitope discovery and mapping: Enables discovery and precise mapping of epitopes for antibody engineering and functional studies.
• Viral antigen analysis: Applicable to characterization of viral proteins, including HIV glycoproteins, for vaccine and therapeutic research.

Methodology:

Integrates structural biology and bioinformatics and employs a fingerprints-based patch model that assesses structural and physicochemical complementarity between antigen patches and mAb CDRs; the model is trained on 860 representative antigen-antibody complexes.