Evaluating Performance of Different RNA Secondary Structure Prediction Programs Using Self-cleaving Ribozymes.

Area of Science:

• Molecular Biology
• Computational Biology
• Bioinformatics

Background:

• Accurate RNA structure identification is crucial for understanding RNA biology and function.
• Various computational, molecular, and chemical methods exist for RNA structure prediction.
• In silico methods offer advantages in speed and cost but often lack accuracy.

Purpose of the Study:

• To compare the accuracy of seven in silico RNA folding prediction tools.
• To evaluate tool performance on self-cleaving ribozyme sequences of varying complexity.
• To identify the most accurate method for predicting biologically relevant RNA secondary structures.

Main Methods:

• Comparative analysis of seven in silico RNA structure prediction tools.
• Testing on dozens of self-cleaving ribozyme sequences.
• Evaluation of prediction accuracy across tasks of differing complexity.

Main Results:

• Many tools performed well on simple RNA folding tasks.
• Significant performance variation was observed for complex RNA folding problems.
• A modern deep learning method demonstrated superior accuracy in complex tasks.

Conclusions:

• Deep learning approaches show promise for accurate RNA secondary structure prediction.
• The tested deep learning method outperformed others in complex folding scenarios.
• This suggests deep learning may be the future of RNA structure prediction algorithms.