cbtope

cbtope predicts conformational B-cell epitopes from antigen amino acid sequences to support identification of B-cell epitope residues for vaccine design.

Key Features:

• Machine learning model: Support Vector Machine (SVM) models were developed for sequence-based prediction of conformational B-cell epitopes.
• Training dataset: Models were trained on a dataset comprising 187 non-redundant protein chains with 2261 antibody-interacting residues of B-cell epitopes.
• Binary Profile of Patterns (BPP): The BPP-based model achieved a maximum Matthews Correlation Coefficient (MCC) of 0.22.
• Physiochemical Profile of Patterns (PPP): The PPP-based approach reached an MCC of 0.17.
• Composition Profile of Patterns (CPP): The CPP-based model achieved an MCC of 0.73 and an accuracy of 86.59%.
• Structure-independent prediction: Provides conformational B-cell epitope predictions from primary amino acid sequences as an alternative to tertiary-structure-dependent methods.

Scientific Applications:

• Epitope identification without structures: Predicts conformational B-cell epitopes when tertiary structures of antigens are unavailable.
• Vaccine design support: Assists in identifying potential B-cell epitope residues in novel or poorly characterized proteins to inform vaccine development.
• Comparative method evaluation: Enables sequence-based benchmarking against structure-based epitope prediction approaches.

Methodology:

Support Vector Machine models were trained on 187 non-redundant protein chains with 2261 antibody-interacting residues using Binary Profile of Patterns (BPP), Physiochemical Profile of Patterns (PPP), and Composition Profile of Patterns (CPP).