Bioactive Peptide Recognition Based on NLP Pre-Train Algorithm.

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Bioactive peptides regulate bodily functions but their identification traditionally relies on time-consuming experiments.
• Existing machine learning models for bioactive peptide recognition are limited by small datasets, impacting performance.

Purpose of the Study:

• To develop an improved computational method for bioactive peptide recognition.
• To enhance the performance of models by utilizing large-scale unlabeled protein sequence data through pre-training.

Main Methods:

• Proposed a novel pre-training method inspired by natural language processing techniques for sequence classification.
• Applied the pre-training method to large-scale protein sequences for enhanced bioactive peptide recognition.

Main Results:

• Achieved superior performance in identifying multiple functional peptides, including anti-cancer, anti-diabetic, anti-hypertensive, anti-inflammatory, and anti-microbial peptides.
• Demonstrated significant improvements in precision (7.2%), coverage (6.9%), accuracy (6.1%), and absolute true (4.2%) compared to advanced models via 5-fold cross-validation.
• Showcased superior prediction performance for single functional peptides, particularly anti-cancer and anti-microbial peptides, which often have longer sequences.

Conclusions:

• The proposed pre-training method effectively enhances bioactive peptide recognition by leveraging extensive unlabeled protein data.
• This computational approach offers a more efficient and accurate alternative to traditional experimental methods for identifying functional peptides.