An efficient algorithm for exploring RNA branching conformations under the nearest-neighbor thermodynamic model

Area of Science:

• Computational Biology
• Bioinformatics
• Molecular Biology

Background:

• The Nearest-Neighbor Thermodynamic Model is standard for RNA secondary structure prediction.
• Multiloop energy is modeled using a linear entropic penalty with three branching parameters.
• Reparameterizing these parameters and exploring alternative branching conformations can improve predictions, but prior methods were computationally inefficient.

Purpose of the Study:

• To develop a novel algorithm for efficient exploration of RNA multiloop parameter space.
• To identify all distinct branching structures optimal under different branching parameters for a given RNA sequence.
• To enable comprehensive evaluation of the structural landscape across parameter choices for improved RNA structure prediction.

Main Methods:

• A novel algorithm partitions the parameter space to identify distinct branching structures.
• The method uses minimal minimum free energy computations.
• Efficiently computes the full parameter-space partition and associated optimal structures.

Main Results:

• The algorithm efficiently computes the full parameter-space partition and optimal structures.
• Application to the Archive II dataset shows substantial improvement potential over default predictions.
• Optimal prediction accuracy is sensitive to auxiliary modeling decisions like treatment of lonely base pairs and dangling ends.

Conclusions:

• Conventional multiloop parameters may limit RNA structure prediction accuracy.
• Exploring alternative parameterizations is tractable and worthwhile for improving predictions.
• The efficient partitioning algorithm makes this exploration feasible for longer sequences and larger datasets.