A New Method of RNA Secondary Structure Prediction Based on Convolutional Neural Network and Dynamic Programming

Area of Science:

• Computational Biology
• Bioinformatics
• Genomics

Background:

• RNA secondary structure is crucial for gene function research.
• Existing minimum free energy (MFE) algorithms struggle with accuracy for longer RNA sequences due to deviations from optimal folding.
• Accurate prediction of RNA secondary structure is essential for understanding RNA and gene function.

Purpose of the Study:

• To develop a novel algorithm for RNA secondary structure prediction that overcomes the limitations of MFE methods, particularly for longer sequences.
• To enhance the accuracy of RNA secondary structure prediction using computational approaches.

Main Methods:

• A convolutional neural network (CNN) model was developed using experimental RNA sequence and structure data.
• The CNN model extracts implicit features to predict base-pairing probabilities.
• An enhanced dynamic programming method was applied to the predicted probabilities to determine the optimal RNA secondary structure.

Main Results:

• The proposed method demonstrated superior performance in predicting secondary structures for three benchmark RNA families.
• The novel algorithm achieved a 30% higher prediction success rate compared to common RNA secondary structure prediction algorithms.
• The accuracy improvement is attributed to the deep learning model's ability to handle complex RNA structures.

Conclusions:

• The combined CNN and dynamic programming approach offers a significant advancement in RNA secondary structure prediction accuracy.
• This method shows great promise for future RNA research, especially as large-scale RNA structure data becomes more available.
• The developed algorithm provides a more reliable tool for computational RNA structure analysis.