Maximum expected accuracy structural neighbors of an RNA secondary structure

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• RNA molecules regulate gene expression and alternative splicing via allosteric mechanisms.
• Accurate prediction of RNA conformational switches is crucial for understanding gene regulation.
• Existing tools struggle to accurately detect general, entire riboswitches (aptamer and expression platform).

Purpose of the Study:

• To develop an accurate algorithm for detecting general RNA conformational switches.
• To improve upon existing RNA structure prediction tools by incorporating Maximum Expected Accuracy (MEA).
• To provide a method for identifying both aptamer and expression platform components of riboswitches.

Main Methods:

• Developed the RNAborMEA algorithm based on approximating Boltzmann probabilities of k-neighbors of a given RNA secondary structure.
• Utilized relative frequency counts over sampled structures to approximate probabilities with rigorous bounds on accuracy and confidence.
• Implemented computation of Maximum Expected Accuracy (MEA) structures for k-neighbors.

Main Results:

• Proved that Boltzmann probabilities of k-neighbors can be approximated using sampling with a provable accuracy bound.
• RNAborMEA computes MEA(k) structures for all k-neighbors, with computation time O(n³ · K²) and memory O(n² · K).
• Applied RNAborMEA to analyze TPP riboswitches and the purine riboswitch class.

Conclusions:

• RNAborMEA provides a novel tool for RNA conformational switch detection, complementing existing methods.
• The algorithm's results are distinct from tools like RNAbor, RNAshapes, and paRNAss, offering orthogonal information.
• Source code for RNAborMEA is publicly available for download.