lbtope

lbtope predicts linear B-cell epitopes from peptide sequences to identify antigenic regions for vaccine design and immunological studies.

Key Features:

• Dataset utilization: Uses experimentally validated data from the Immune Epitope Database (IEDB), including Lbtope_Variable (14,876 epitopes; 23,321 non-epitopes), Lbtope_Fixed Length (12,063 epitopes; 20,589 non-epitopes), and Lbtope_Confirm (1,042 epitopes; 1,795 non-epitopes).
• Negative controls: Employs experimentally validated non-B-cell epitopes as negative examples instead of random peptides.
• Machine learning models: Implements Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) classifiers.
• Feature representations: Uses binary profiles, dipeptide compositions, and amino acid pair (AAP) profiles as input features.
• Performance evaluation: Models were evaluated after removing highly identical peptides, reporting accuracies ranging approximately from 54% to 86%.

Scientific Applications:

• Vaccine development: Provides epitope predictions to prioritize antigenic peptides for vaccine candidate selection.
• Immunology research: Assists studies of B-cell responses by identifying linear antigenic regions in proteins.
• Peptide-based vaccine design: Informs design of peptide-based vaccines through identification of predicted linear B-cell epitopes.
• Personalized medicine: Contributes to personalized immunogen design by improving precision of epitope prediction.

Methodology:

Training and evaluation used IEDB-derived datasets (Lbtope_Variable, Lbtope_Fixed Length, Lbtope_Confirm) with experimentally validated non-B-cell epitopes as negatives; features included binary profiles, dipeptide composition, and amino acid pair (AAP) profiles, and models (SVM, KNN) were evaluated after removing highly identical peptides, yielding accuracies of ~54%–86%.