iBitter-Stack: A multi-representation ensemble learning model for accurate bitter peptide identification.

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Bitter peptides are important in food science and drug discovery.
• Experimental identification of bitter peptides is costly and time-consuming.
• Efficient computational methods are needed for bitter peptide classification.

Purpose of the Study:

• To develop a novel stacking-based ensemble learning framework for accurate bitter peptide classification.
• To enhance the reliability and efficiency of identifying bitter peptides computationally.
• To provide a user-friendly tool for real-time bitter peptide screening.

Main Methods:

• Integrated diverse sequence-based feature representations.
• Employed a two-layer stacking ensemble learning framework.
• Utilized logistic regression in the second layer for refined predictions.

Main Results:

• Achieved 96.09% accuracy and 0.9220 MCC on an independent test set.
• Demonstrated superior performance compared to existing predictive methods.
• Developed a freely accessible web server for real-time bitter peptide screening.

Conclusions:

• The proposed stacking ensemble model offers a robust and reliable computational tool for bitter peptide identification.
• The developed web server enhances accessibility for researchers and practitioners.
• This approach addresses the growing need for efficient bitter peptide prediction in the post-genomic era.