ABCpred

ABCpred predicts linear B cell epitope regions within antigen sequences to support vaccine design, disease diagnostics, and allergy research.

Key Features:

• Artificial Neural Networks: ABCpred employs feed-forward neural networks (FNN) and Jordan recurrent neural networks (RNN) for epitope prediction.
• Training Dataset: Networks were trained on 700 non-redundant B-cell epitopes from the Bcipep database and 700 non-epitopes randomly selected from Swiss-Prot.
• Optimized Parameters: A Jordan RNN with a single hidden layer of 35 units and an input window length of 16 achieved maximum accuracy of 65.93% under fivefold cross-validation.
• Performance Metrics: Reported performance includes sensitivity 67.14%, specificity 64.71%, and positive predictive value 65.61%.

Scientific Applications:

• Vaccine Development: Predicts B cell epitopes to aid selection of synthetic vaccine candidates.
• Disease Diagnosis and Allergy Research: Identifies epitope regions to inform diagnostic marker discovery and studies of allergic immune responses.

Methodology:

ABCpred trains FNN and Jordan RNN models on the Bcipep and Swiss-Prot datasets, evaluates performance by fivefold cross-validation, and optimizes network architecture and training parameters (hidden layer size and input window length) to maximize predictive accuracy.