IUP-BERT: Identification of Umami Peptides Based on BERT Features

Area of Science:

• Food Science
• Biotechnology
• Computational Biology

Background:

• Umami taste is crucial for food palatability, with umami peptides being key contributors.
• Traditional laboratory methods for identifying umami peptides are inefficient for large-scale screening.

Purpose of the Study:

• To develop a rapid and accurate computational method for predicting umami peptides based on their amino acid sequences.
• To introduce iUP-BERT, a novel predictor leveraging deep learning for umami peptide identification.

Main Methods:

• Utilized BERT (bidirectional encoder representations from transformer) for deep learning-based feature extraction from peptide sequences.
• Employed synthetic minority over-sampling technique (SMOTE) and support vector machine (SVM) for model construction and prediction.
• Developed a web server for open access to the iUP-BERT prediction tool.

Main Results:

• The iUP-BERT model demonstrated superior performance compared to existing methods in predicting umami peptides.
• Extensive validation through cross-validation and independent testing confirmed the model's effectiveness and robustness.
• Achieved high accuracy in predicting probabilistic scores for potential umami peptides.

Conclusions:

• iUP-BERT represents a significant advancement in sequence-based umami peptide prediction.
• The developed tool offers an efficient alternative to traditional methods, facilitating faster discovery of umami peptides.
• This predictor has the potential to aid research in enhancing the flavor profiles of food products and dietary supplements.