When a ribonucleic acid (RNA) molecule folds, it often does not adopt a single, well-defined conformation. The folding energy landscape of an RNA is highly dependent on its nucleotide sequence and molecular environment. Cellular molecules sometimes alter the energy landscape, thereby changing the ensemble of likely low-energy conformations. The effects of these energy landscape changes on the conformational ensemble are particularly challenging to visualize for large RNAs. We have created a robust approach for visualizing the conformational ensemble of RNAs that is well suited for in vitro versus in vivo comparisons. Our method creates a stable map of conformational space for a given RNA sequence. We first identify single point mutations in the RNA that maximally sample suboptimal conformational space based on the ensemble’s partition function. Then, we cluster these diverse ensembles to identify the most diverse partition functions for Boltzmann stochastic sampling. By using, to our knowledge, a novel nestedness distance metric, we iteratively add mutant suboptimal ensembles to converge on a stable 2D map of conformational space. We then compute the selective 2′ hydroxyl acylation by primer extension (SHAPE)-directed ensemble for the RNA folding under different conditions, and we project these ensembles on the map to visualize. To validate our approach, we established a conformational map of the Vibrio vulnificus add adenine riboswitch that reveals five classes of structures. In the presence of adenine, projection of the SHAPE-directed sampling correctly identified the on-conformation; without the ligand, only off-conformations were visualized. We also collected the whole-transcript in vitro and in vivo SHAPE-MaP for human β-actin messenger RNA that revealed similar global folds in both conditions. Nonetheless, a comparison of in vitro and in vivo data revealed that specific regions exhibited significantly different SHAPE-MaP profiles indicative of structural rearrangements, including rearrangement consistent with binding of the zipcode protein in a region distal to the stop codon.

The precise impact of inherited and somatic mutations on messenger RNA (mRNA) structure remains poorly understood. Recent technological advances that leverage next generation sequencing to obtain experimental structure data, such as SHAPE-MaP (Selective 2' Hydroxyl Acylation by Primer Extension and Mutational Profiling), can reveal structural effects of mutations especially when this data is rigorously incorporated in RNA structural modeling. Here we analyze the ability of SHAPE-MaP to detect the relatively subtle structural changes caused by single nucleotide mutations. We find that allele-specific sorting greatly improved the detection ability of SHAPE-MaP. Thus, we used SHAPE-MaP with a novel combination of clone-free robotic mutagenesis and allele-specific sorting to perform a rapid, comprehensive survey of non-coding somatic and inherited riboSNitches in two cancer-associated mRNAs, TPT1 and LCP1. Combined with rigorous thermodynamic modeling of the Boltzmann suboptimal ensemble we identified a subset of somatic and Lackey 2 inherited mutations that change TPT1 and LCP1 RNA structure, with approximately 14% of all variants identified acting as riboSNitches. To confirm that these in vitro structures were biologically relevant, we tested how dependent TPT1 and LCP1 mRNA structures are on their environments. We performed SHAPE-MaP on TPT1 and LCP1 mRNAs in the presence or absence of cellular proteins and found that both TPT1 and LCP1 have similar overall folds in all conditions. RiboSNitches identified within these mRNAs in vitro are likely to exist under biological conditions. Overall, these data reveal a remarkably robust mRNA structural landscape where differences in environmental conditions and most sequence variants do not significantly alter RNA structural ensembles. Finally, predicting riboSNitches in mRNAs from sequence alone remains particularly challenging; these data will provide the community with a benchmark for further algorithmic development.

MotivationMutations (or Single Nucleotide Variants) in folded RiboNucleic Acid structures that cause local or global conformational change are riboSNitches. Predicting riboSNitches is challenging, as it requires making two, albeit related, structure predictions. The data most often used to experimentally validate riboSNitch predictions is Selective 2′ Hydroxyl Acylation by Primer Extension, or SHAPE. Experimentally establishing a riboSNitch requires the quantitative comparison of two SHAPE traces: wild-type (WT) and mutant. Historically, SHAPE data was collected on electropherograms and change in structure was evaluated by ‘gel gazing.’ SHAPE data is now routinely collected with next generation sequencing and/or capillary sequencers. We aim to establish a classifier capable of simulating human ‘gazing’ by identifying features of the SHAPE profile that human experts agree ‘looks’ like a riboSNitch.ResultsWe find strong quantitative agreement between experts when RNA scientists ‘gaze’ at SHAPE data and identify riboSNitches. We identify dynamic time warping and seven other features predictive of the human consensus. The classSNitch classifier reported here accurately reproduces human consensus for 167 mutant/WT comparisons with an Area Under the Curve (AUC) above 0.8. When we analyze 2019 mutant traces for 17 different RNAs, we find that features of the WT SHAPE reactivity allow us to improve thermodynamic structure predictions of riboSNitches. This is significant, as accurate RNA structural analysis and prediction is likely to become an important aspect of precision medicine.Availability and ImplementationThe classSNitch R package is freely available at http://classsnitch.r-forge.r-project.org.Supplementary information Supplementary data are available at Bioinformatics online.

No abstract