B i o A I L a b

Welcome To BioAi−Lab

Introduction

Accurate identification of bitter peptides is essential for bitter peptide research. Machine learning has been widely used to assess the properties of peptides. Although models using sequence information have evolved in the context of bitter peptides, there is still room for improvement in their predictive performance. In the present study, we introduced a novel predictive tool, iBitter-GRE, designed to improve the accuracy of bitter peptide identification. Our model uses ESM-2 and traditional descriptors that capture the physical and biochemical properties of bitter peptides for feature extraction. To expand the model's learning capabilities, we adopted a stacking approach to integrate multiple learners. Feature contributions were analyzed using SHAP values. Validation by domain experts confirmed that our model effectively identifies the key biochemical characteristics of bitter peptides. Benchmark experiments, including 10-fold cross-validation and independent tests, showed that iBitter-GRE achieves higher accuracy than existing methods. We believe that iBitter-GRE is a valuable tool for the discovery and identification of bitter peptides.



Framework