PLM-IL4: Enhancing IL-4-inducing peptide prediction with protein language model

Area of Science:

• Immunology
• Bioinformatics
• Computational Biology

Background:

• Interleukin-4 (IL-4) plays a critical role in immune regulation and allergic responses.
• Predicting IL-4 inducing peptides is vital for advancing immunotherapy and vaccine development.
• Existing prediction methods face challenges with data imbalance and feature extraction.

Purpose of the Study:

• To improve the predictive accuracy of IL-4-inducing peptides.
• To address data imbalance issues in peptide datasets.
• To enhance deep feature extraction for better peptide prediction.

Main Methods:

• Utilized SMOTE (Synthetic Minority Over-sampling Technique) and ENN (Edited Nearest Neighbors) for dataset balancing.
• Employed a 30-layer ESM-2 model for deep feature extraction from peptide sequences.
• Applied a hyperparameter-tuned Gated Recurrent Unit (GRU) model for classification.

Main Results:

• Achieved a high Area Under the Curve (AUC) of 0.98.
• Reached a prediction accuracy of 93.1%.
• Demonstrated significant improvements in IL-4 inducing peptide prediction.

Conclusions:

• The developed method offers enhanced predictive accuracy for IL-4-inducing peptides.
• This approach holds significant potential for future immunotherapy and vaccine design.
• The PLM-IL4 web server and datasets are publicly available for research.