A BERT-based rice enhancer identification model combined with sequence-representation differential entropy...

Area of Science:

• Genomics
• Molecular Biology
• Bioinformatics

Background:

• Gene expression regulation is vital for rice molecular breeding and yield improvement.
• Identifying enhancers, key gene expression regulators, is challenging in functional genomics.
• Current deep learning methods for rice enhancer identification lack efficient feature extraction and generalization.

Purpose of the Study:

• To introduce a novel model, RiceEN-BERT-SVM, for efficient and accurate rice enhancer identification.
• To integrate DNABERT-2 for feature extraction and Support Vector Machine (SVM) for classification.
• To elucidate the mechanism of performance optimization using differential entropy analysis.

Main Methods:

• Developed the RiceEN-BERT-SVM model integrating DNABERT-2 and SVM.
• Utilized differential entropy analysis to interpret feature representations and model performance.
• Conducted 5-fold cross-validation and independent testing to evaluate accuracy.

Main Results:

• RiceEN-BERT-SVM achieved 88.05% accuracy in 5-fold cross-validation and 87.55% in independent testing, surpassing SOTA models.
• Fine-tuning further improved performance, reaching a final accuracy of 93.63%.
• Differential entropy analysis showed increasing separation of positive and negative sample features with fine-tuning, correlating with performance gains.

Conclusions:

• The RiceEN-BERT-SVM model offers an efficient and accurate tool for rice enhancer identification.
• Differential entropy analysis provides a visually interpretable framework for optimizing biological sequence analysis models.
• This approach enhances understanding of gene expression regulation for improved rice breeding and yield.