RNAG: a new Gibbs sampler for predicting RNA secondary structure for unaligned sequences

Area of Science:

• Bioinformatics
• Computational Biology
• Molecular Biology

Background:

• RNA secondary structure is crucial for RNA function and interactions.
• Predicting RNA secondary structures for families of unaligned sequences is a significant challenge.
• Existing methods for RNA structure prediction have limitations.

Purpose of the Study:

• To introduce RNAG, a novel global structural alignment algorithm for predicting consensus secondary structures of unaligned RNA sequences.
• To leverage a blocked Gibbs sampling approach for enhanced convergence time and prediction accuracy.
• To characterize the uncertainty associated with RNA structure predictions.

Main Methods:

• Development of the RNAG algorithm, employing a blocked Gibbs sampling strategy.
• Iterative sampling from conditional probability distributions P(Structure | Alignment) and P(Alignment | Structure).
• Utilizing sampled structures to analyze the posterior space and assess prediction uncertainty.

Main Results:

• RNAG demonstrated substantial improvement in RNA structure prediction compared to existing methods across three public datasets.
• Analysis of 17 RNA families revealed compact sampled structures around ensemble centroids.
• At least 11 RNA families exhibited two or more well-separated clusters of predicted structures, indicating distinct structural possibilities.

Conclusions:

• RNAG provides a powerful new tool for predicting consensus secondary structures of RNA families.
• The algorithm effectively addresses the challenge of uncertainty in high-dimensional RNA structure prediction.
• The method offers significant advancements over extant prediction techniques, aiding in understanding RNA function.