MsipNet: a multi-scale representation learning framework for predicting protein-RNA interaction.

Area of Science:

• Molecular Biology
• Bioinformatics
• Computational Biology

Background:

• Protein-RNA interactions (PRIs) are crucial for post-transcriptional gene regulation, affecting RNA splicing, stability, and translation.
• Understanding PRIs is vital for elucidating gene regulatory networks and disease mechanisms linked to mutations.
• Accurate PRI identification bridges basic research and biomedical applications.

Purpose of the Study:

• To develop an advanced computational framework for predicting protein-RNA interactions.
• To improve the accuracy and efficiency of PRI prediction using a multimodal learning strategy.
• To provide a robust tool for prioritizing functional mutations and advancing mechanistic studies.

Main Methods:

• Introduced MsipNet, a multi-scale representation learning framework.
• Integrated global and local RNA sequence features with structural information.
• Employed a hybrid architecture combining Long Short-Term Memory (LSTM) networks and U-shaped convolution-dilated convolution (UCDC) modules.

Main Results:

• MsipNet outperformed eight state-of-the-art methods across 42 RNA-binding proteins (RBPs) from six cell lines.
• Demonstrated superior performance in predicting binding preferences and identifying biologically validated binding motifs.
• Showcased strong generalizability on unseen data, maintaining high computational efficiency.

Conclusions:

• MsipNet is a robust and interpretable tool for PRI prediction.
• The framework has broad potential for mechanistic studies and biomedical applications, including functional mutation prioritization.
• MsipNet advances the field of computational biology for understanding gene regulation and disease.