Deciphering Complex RNA structures by probing and interactions – PaRNAssus

Secondary structure prediction is an essential step for the de-novo modeling of the architectures of RiboNucleic Acids (RNAs). Advances in high-throughput structural probing of RNA molecules have led to the development of experimentally assisted in-silico prediction methods, that achieve very good agreement with known structures in terms of correctly predicted base pairs. However, the majority of available modeling tools confine themselves to a single particular strategy to incorporate experimental data into the prediction algorithms. Thereby, they mostly neglect the differences in the probing reagents mode of interaction, and their efficiency under different experimental conditions. Moreover, very few methods allow for the simultaneous analysis of multiple probing data, or the prediction of nucleotide interactions beyond the classic secondary structure definition, thereby neglecting important structural features such as pseudoknots, or tertiary structure motifs.

In this project, we propose to establish new hybrid, both experimental and in silico, protocols to assist structure modeling based on new data. By closely investigating various chemical probing reagents, we will assess their potential to derive complementary information from multiple experiments under different conditions, and using different reagents. This data will then be refined to best fit into existing RNA secondary structure prediction approaches, and novel heuristics for pseudoknot and non-canonical structure prediction. To strengthen such approaches, we will probe a collection of recurring tertiary motifs, and will use statistical machine learning as well as molecular dynamics, to define discriminative probing patterns.
A capacity to adopt multiple folds is a physiological property of many functional RNAs. However most biophysical methods work under the assumption of a single conformations, despite the fact that probing methods produce a signal which is averaged over existing conformations. This is currently one of the key obstacles to the accurate modeling from structure probing data. To bypass this problem and thus deconvolute the signal, we propose to develop an innovative workflow which consists in probing at once an ensemble of mutants of the RNA of interest.
Finally, `many RNAs are known to fold co-transcriptionally, therefore folding kinetics are crucial to understand which states are effectively populated in the lifetime of an RNA. In this project to propose an original experimental strategy combining high-speed probing and high throughput sequencing to detect folding intermediates during transcription.

Our project aims at providing the community with portable experimental and analytical workflows, allowing biologists to rapidly model accurate and relevant structures for their RNAs of interest. Different workflows will be defined according the RNA type and the level of structural details desired by the experimentalist. We will deliver mature implementations, and web based software for each contributed method. It builds on two past projects funded by ANR/FWF and FRM respectively, and will be carried out by a consortium harnessing a wide combination of expertise and skills. This multidisciplinary project brings together RNA researchers from various disciplines, including computer science, biochemistry, structural biology, mathematics and physics, to help uncover the true nature of structured RNAs.