TurboFold: iterative probabilistic estimation of secondary structures for multiple RNA sequences

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• RNA secondary structure prediction is crucial for understanding RNA function.
• Accurate methods often involve homologous sequences but are computationally intensive.
• TurboFold offers an efficient alternative for multi-sequence RNA structure prediction.

Purpose of the Study:

• To present TurboFold, a novel and efficient method for predicting secondary structures of multiple RNA sequences.
• To improve the accuracy of RNA secondary structure prediction by combining intrinsic and extrinsic information.
• To provide a computationally feasible approach for analyzing homologous RNA sequences.

Main Methods:

• TurboFold uses an iterative probabilistic approach combining intrinsic thermodynamic nearest-neighbor models with extrinsic information from homologous sequences.
• Extrinsic information is derived from pairwise sequence alignments and probabilities of base pairing from previous iterations.
• The method incorporates free energy modifications into a partition function computation.

Main Results:

• TurboFold demonstrates improved base pairing probability estimates with iterative refinement, with significant gains within three iterations.
• TurboFold-MEA achieves accuracy comparable to leading secondary structure prediction methods.
• The method exhibits favorable computational and memory scaling for sequence length and number of sequences.

Conclusions:

• TurboFold is an effective iterative method for predicting RNA secondary structures, efficiently integrating comparative and thermodynamic data.
• It accurately predicts structures for significantly diverged homologous sequences without enforcing strict structural commonality.
• TurboFold is available as part of the RNAstructure package.