DeepBCE

DeepBCE predicts immunogenic B-cell epitopes (BCEs) from proteomics sequences using sequence-based features and convolutional neural networks (CNNs) to support epitope identification for vaccine, antibody, and therapeutic development.

Key Features:

• Target: Predicts immunogenic B-cell epitopes (BCEs), which interact with antigens to initiate immune responses.
• Input data: Uses proteomics sequences and sequence-based features as model inputs.
• Primary model: Implements a convolutional neural network (CNN)-based model as the core predictive architecture.
• Model repertoire: Employs both classical machine learning models and deep learning models (DLMs).
• Comparative performance: The CNN-based model was demonstrated to outperform other existing models in predictive accuracy.
• Performance metrics: Reported accuracy (ACC) 0.878, F-measure 0.871, and area under the ROC curve (AUC) 0.945.
• MCC improvement: Achieves an average improvement of 58.7% based on Matthews Correlation Coefficient (MCC) relative to comparator methods.

Scientific Applications:

• Vaccine development: Identification of immunogenic BCEs for antigen selection and vaccine antigen design.
• Antibody and therapeutic development: Epitope prediction to inform antibody target selection and therapeutic design.
• Antigen–antibody interaction studies: Prioritization of epitope candidates for experimental validation of antigen–antibody binding.
• Biotechnology and biomedicine: Use in studies requiring mapping of immunity-stimulating B-cell epitopes.

Methodology:

Predicts BCEs from proteomics sequences using sequence-based features and applies both classical machine learning and deep learning models, with a CNN-based architecture as the core predictive method.