Sep 02 2015 SHAPE directed RNA folding with the ViennaRNA Package

Bridging Experiment with Computation

Modern experiments generate SHAPE reactivity profiles through selective 2’-hydroxyl acylation analyzed by primer extension, a method that reveals nucleotide flexibility in vitro or in vivo. To translate these reactivities into structural guidance, the Deigan, Zarringhalam, and Washietl methods each adjust the energy parameters used in traditional dynamic programming: Deigan’s model applies pseudo-energies directly to loops, Zarringhalam’s method uses reactivity-specific weights, and Washietl’s approach infers perturbation of the energy parameters to optimally fit the SHAPE signal. Together, these innovations allow the prediction of both the minimum free energy (MFE) structure and the ensemble of probable folds with improved sensitivity to experimental data.

Performance and Benchmarking

To validate these implementations, we benchmarked against a curated set of RNAs with known reference structures, analyzing metrics like sensitivity and positive predictive value for predicted base pairs. The comparisons covered both the MFE structure and the full partition function ensemble, demonstrating that SHAPE-guided predictions outperform purely thermodynamic models, especially for challenging motifs. Detailed supplementary data outline our benchmark strategies, energy parameter settings, and the impact on diverse RNA molecules, confirming the value of integrating experimental evidence into computational pipelines.

Getting Started: Command Line and Source Code

Users can install binary packages for several Linux distributions or build from source code available on our repository. The command line interface supports commands for folding, partition function calculation, and SHAPE-guided modes, with options to specify and adjust energy parameters directly. Whether you’re scripting large-scale screens or running individual analyses, our tools and comprehensive documentation provide everything needed to apply dynamic programming algorithms to your nucleic acid research.

For a deep dive, don’t miss the Supplementary Data, which includes extensive background on each method, parameter details, and full benchmark results for our SHAPE-integrated algorithms. Dive into the code, explore the options, and see how experimental data can guide your next discovery in RNA structure prediction.