BPFun: a deep learning framework for bioactive peptide function prediction using multi-label strategy by...

Area of Science:

• Bioactive peptides research
• Computational biology
• Bioinformatics

Background:

• Bioactive peptides exhibit diverse physiological effects, necessitating accurate identification of their multiple functions.
• Traditional experimental methods for identifying bioactive peptide functions are resource-intensive and time-consuming.
• Developing computational approaches is crucial for efficient and accurate prediction of bioactive peptide functions.

Purpose of the Study:

• To propose a novel deep learning model, BPFun, for predicting multiple functions of bioactive peptides.
• To enhance the accuracy and efficiency of bioactive peptide function prediction compared to traditional methods.
• To address the challenge of data imbalance in bioactive peptide datasets.

Main Methods:

• Utilized a deep learning approach, BPFun, incorporating biological and physicochemical features of bioactive peptides.
• Employed data augmentation techniques to mitigate data imbalance issues.
• Combined multi-scale convolutional networks, Bi-LSTM layers, and a self-attention mechanism for feature extraction and fusion.

Main Results:

• BPFun accurately predicted seven diverse functions of bioactive peptides, including anticancer, antibacterial, and antihypertensive.
• The model achieved an accuracy of 0.6577 and an absolute truth value of 0.6573 on a seven-functional classification dataset.
• BPFun demonstrated superior performance over existing methods in predicting bioactive peptide functions.

Conclusions:

• BPFun provides an effective computational tool for predicting multiple bioactive peptide functions.
• The deep learning architecture, integrating various feature types and attention mechanisms, significantly improves prediction accuracy.
• The developed model offers a valuable resource for accelerating research in bioactive peptides and their applications.