MRM-BERT: a novel deep neural network predictor of multiple RNA modifications by fusing BERT representation and...

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• RNA modifications are vital for biological processes and disease.
• Accurate identification of RNA modification sites is crucial for functional studies.

Purpose of the Study:

• To develop a unified deep learning framework for predicting multiple RNA modification types.
• To enhance the accuracy and efficiency of RNA modification site prediction.

Main Methods:

• Developed MRM-BERT, a hybrid deep learning model.
• Integrated pre-trained DNABERT sequence representation with CNNs and traditional sequence features.
• Evaluated performance on 12 common RNA modifications (e.g., m6A, m5C, m1A).

Main Results:

• MRM-BERT achieved superior prediction performance across all 12 RNA modification types.
• The hybrid approach demonstrated higher accuracy compared to existing models, measured by AUC.
• The model is accessible as an online tool and open-source code.

Conclusions:

• MRM-BERT offers an effective and efficient solution for predicting multiple RNA modification sites.
• This advancement contributes to a deeper understanding of RNA biology.
• The tool supports the development of novel RNA-targeted therapeutic strategies.