Accurate Classification of RNA Structures Using Topological Fingerprints

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• Functionally similar RNAs can exhibit low sequence similarity, posing challenges for classification.
• RNA structures are complex, often containing pseudoknots and variations in stem size and spacing.
• Existing methods may struggle with incomplete or predicted RNA structures.

Purpose of the Study:

• To develop a topological approach for identifying and classifying functionally related RNAs.
• To create an RNA fingerprint based on topological spectra for robust RNA comparison.
• To address limitations of sequence-based methods and incomplete structural data.

Main Methods:

• Representing RNA structures as graphs to generate topological spectra (RNA fingerprints).
• Comparing RNA fingerprints (sets of subgraphs) to identify similarities.
• Evaluating the method on curated RNA families with diverse structures, including pseudoknots.

Main Results:

• The RNA fingerprint approach successfully identifies functionally related RNAs with low sequence similarity (ROC AUC > 0.95).
• The method is tolerant to significant omissions (up to 30%) in RNA stems, indicating a need for incomplete structures.
• Inclusion of pseudoknots broadens applicability compared to secondary structure-only methods.

Conclusions:

• Topological analysis provides a robust method for RNA classification, overcoming sequence and structural variability.
• The RNA fingerprint approach is effective even with incomplete or predicted RNA structures.
• This method enhances the ability to discover and classify functionally related RNAs across diverse families.