Accurate prediction of toxicity peptide and its function using multi-view tensor learning and latent semantic...

Area of Science:

• Computational biology
• Pharmacology
• Bioinformatics

Background:

• Therapeutic peptides are vital in disease treatment and drug discovery.
• Peptide toxicity poses a significant challenge in peptide drug therapy.
• Efficient computational methods are needed to identify toxic peptides.

Purpose of the Study:

• To develop a computational method for predicting peptide toxicity.
• To classify toxicity peptides into multi-functional types.
• To address the challenge of identifying toxic peptides in the post-genomics era.

Main Methods:

• Developed a two-level predictor, ToxPre-2L, using multi-view tensor learning and latent semantic learning.
• Employed multi-label learning with feature-induced labels to manage information redundancy.
• Utilized low-rank constraint learning to capture correlations among multi-labels.

Main Results:

• ToxPre-2L demonstrated superior performance compared to existing computational methods.
• The predictor accurately identifies toxicity peptides.
• Effectively classifies multi-functional toxicity peptides.

Conclusions:

• ToxPre-2L offers an effective computational approach for predicting peptide toxicity.
• The method aids in identifying and classifying multi-functional toxic peptides.
• Enhances the safety and development of therapeutic peptides.