Rm-LR: A long-range-based deep learning model for predicting multiple types of RNA modifications.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• RNA post-transcriptional modifications are crucial for regulating gene expression and function.
• Accurate identification of RNA modification sites is essential for understanding RNA biology.

Purpose of the Study:

• To develop a novel and accurate computational method for predicting multiple types of RNA modifications.
• To leverage deep learning and RNA language models for enhanced prediction accuracy.

Main Methods:

• Proposed Rm-LR method utilizing a long-range-based deep learning approach.
• Incorporated two large-scale RNA language pre-trained models to capture sequential and local features.
• Integrated features using a bilinear attention network for prediction.

Main Results:

• Rm-LR accurately predicts ten RNA modification types (m⁶A, m¹A, m⁵C, m⁵U, m⁶Am, Ψ, Am, Cm, Gm, Um) using only RNA sequences.
• Significantly outperformed state-of-the-art methods on benchmark datasets.
• Demonstrated strong adaptability and robustness across various RNA modifications.

Conclusions:

• Rm-LR provides an effective and superior computational model for RNA modification prediction.
• RNA language pre-trained models enhance the learning of biological sequential representations and model interpretability.
• This work advances accurate and reliable prediction of RNA modifications, offering insights into their complex landscape.