PrMFTP: Multi-functional therapeutic peptides prediction based on multi-head self-attention mechanism and class...

Area of Science:

• Computational Biology
• Drug Discovery
• Bioinformatics

Background:

• Therapeutic peptide prediction is crucial for drug discovery.
• Existing methods primarily focus on mono-functional peptides, neglecting the growing number of multi-functional therapeutic peptides (MFTP).
• There is a need for advanced computational approaches to facilitate MFTP discovery.

Purpose of the Study:

• To propose a novel computational model, PrMFTP, for the accurate prediction of multi-functional therapeutic peptides (MFTP).
• To address the challenge of imbalanced data in MFTP prediction using a class weight optimization scheme.

Main Methods:

• Developed PrMFTP, a model integrating multi-scale convolutional neural networks (CNN), bi-directional long short-term memory (BiLSTM), and multi-head self-attention mechanisms.
• Employed a class weight optimization algorithm to handle label imbalanced data.
• Utilized comprehensive evaluations to compare PrMFTP against state-of-the-art methods.

Main Results:

• PrMFTP effectively extracts and learns informative features from peptide sequences for MFTP prediction.
• The proposed class weight optimization scheme successfully addresses data imbalance issues.
• PrMFTP demonstrated superior performance compared to existing state-of-the-art computational methods for MFTP prediction.

Conclusions:

• PrMFTP represents a significant advancement in computational prediction of multi-functional therapeutic peptides.
• The model's superior performance and user-friendly web server will aid researchers in accelerating therapeutic peptide discovery.
• This work facilitates the identification of novel therapeutic drugs by improving MFTP discovery.