A divide-and-conquer approach based on deep learning for long RNA secondary structure prediction: Focus on...

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• Accurate RNA secondary structure prediction, especially pseudoknots, is vital for understanding RNA function.
• Existing methods struggle with computational complexity and precision for long RNA sequences and pseudoknots.

Purpose of the Study:

• To develop a scalable and precise deep learning method for predicting RNA secondary structures, including pseudoknots, in long RNA sequences.
• To overcome the limitations of current computational approaches for complex RNA structures.

Main Methods:

• A novel divide-and-conquer deep learning strategy named DivideFold.
• Recursive partitioning of long RNA sequences into smaller fragments.
• Integration with existing models for pseudoknot prediction in manageable segments.

Main Results:

• DivideFold demonstrates superior performance in predicting secondary structures with pseudoknots for long RNA sequences.
• The method effectively scales to handle long RNA sequences, a significant challenge for prior approaches.
• Outperforms state-of-the-art methods in both pseudoknot and comprehensive secondary structure prediction.

Conclusions:

• DivideFold offers a robust and scalable solution for predicting RNA secondary structures, including pseudoknots, in long sequences.
• The approach enhances our ability to study RNA folding and function through accurate structural predictions.
• Provides a valuable tool for researchers in RNA biology and bioinformatics.