BepiPred

BepiPred predicts linear B-cell epitopes in protein sequences to identify antibody-recognized sites for applications such as vaccine design and diagnostic test development.

Key Features:

• Predictive target: Identifies linear B-cell epitopes (antibody-recognized sites) in protein sequences.
• Algorithmic integration: Combines a hidden Markov model with a propensity scale method to generate predictions.
• Training datasets: Uses three datasets of linear B-cell epitope-annotated proteins compiled from literature, the AntiJen database, and the Los Alamos HIV database.
• Validation strategy: Employs an unbiased validation approach by testing predictive methods on datasets not used for training.
• Performance evaluation: Assesses methods using non-parametric measures and Receiver Operating Characteristic (ROC) curves.
• Method performance: Identified the hidden Markov model as the most effective single method among those tested.

Scientific Applications:

• Vaccine development: Identification of candidate linear B-cell epitopes for inclusion in vaccine antigen design.
• Diagnostic test design: Mapping antibody-binding sites to inform diagnostic reagent selection and epitope-based assays.
• Immunological research: Epitope mapping in proteins, including HIV proteins, to study antibody recognition.

Methodology:

Three datasets of linear B-cell epitope-annotated proteins were constructed from literature, the AntiJen database, and the Los Alamos HIV database; predictive methods were evaluated using an unbiased validation approach on datasets not used for training, assessed with non-parametric measures and ROC curves, and the final BepiPred method integrates a hidden Markov model with a propensity scale method.