DeepBP: Ensemble deep learning strategy for bioactive peptide prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Peptide Science

Background:

• Bioactive peptides are crucial molecules regulating physiological processes, immune responses, and exhibiting antibacterial effects.
• Their significant roles drive applications in drug development, food science, and biotechnology.
• Understanding peptide mechanisms is key for novel drug discovery and disease treatment.

Purpose of the Study:

• To develop an accurate and efficient method for predicting bioactive peptides.
• To enhance the prediction of angiotensin-converting enzyme (ACE) inhibitory peptides and anticancer peptides (ACPs).
• To leverage ensemble learning for improved peptide prediction performance.

Main Methods:

• Utilized protein language model-evolutionary scale modeling (ESM-2) for feature extraction.
• Employed generative adversarial capsule networks (CapsuleGAN), gated recurrent units (GRU), and convolutional neural networks (CNN) as base classifiers.
• Implemented ensemble learning through a weighted voting method based on individual model accuracy.

Main Results:

• Achieved high prediction accuracy on ACE inhibitory peptide dataset (balanced accuracy 0.926, MCC 0.831, AUC 0.966).
• Demonstrated strong performance on the anticancer peptide (ACP) dataset (ACC 0.779, MCC 0.558).
• The ensemble model significantly outperformed existing methods on both datasets.

Conclusions:

• Ensemble learning with CapsuleGAN, GRU, and CNN effectively predicts functional peptides.
• The developed method offers a significant advancement in accurately and rapidly identifying ACE inhibitory peptides and ACPs.
• This work provides valuable insights for predicting other types of functional peptides, with code and data publicly available.