When does additional information improve accuracy of RNA secondary structure prediction?

Area of Science:

• Computational Biology
• Bioinformatics
• Structural Biology

Background:

• RNA secondary structure is vital for biological function.
• Accurate prediction of RNA secondary structure remains a challenge in computational biology.
• Auxiliary information can enhance secondary structure prediction accuracy.

Purpose of the Study:

• To investigate features derived from suboptimal RNA structures for improved prediction accuracy.
• To introduce and evaluate 'profiles' as a similarity measure for competing substructures.
• To explore the utility of topological data analysis, specifically persistence landscapes, for RNA structure prediction.

Main Methods:

• Development of an n-dimensional representation for RNA substructure profiles.
• Application of topological data analysis (persistence landscapes) to extract topological features.
• Construction of random forest classifiers utilizing similarity (profile) and topological features.
• Extensive testing on two RNA sequence datasets to assess classification accuracy and feature importance.

Main Results:

• Similarity features (profiles) are more impactful for classifying sequences with similar structures.
• Topological features derived from persistence landscapes are more important for classifying sequences with dissimilar structures.
• Demonstrated the effectiveness of novel features in improving RNA secondary structure prediction.

Conclusions:

• Novel similarity and topological features enhance RNA secondary structure prediction.
• The choice of features depends on the structural similarity of the training sequences.
• This work provides new tools and insights for computational RNA structure analysis.