Year: 2012
Type: dissertation
Abstract: In the past decade, there has been an explosion in the number of identified regulatory non-protein-coding RNAs (ncRNAs). Famous representatives are microRNAs in eukaryotes and small RNAs (sRNAs) in bacteria, which both act as post-transcriptional regulators by base pairing with messenger RNAs (mRNAs). For the functional characterisation of these base-pairing ncRNAs, an understanding of the principles that govern the RNA-RNA interaction formation is as essential as is the availability of large-scale target identification approaches. Therefore, I addressed these aspects by means of computational methods and analyses. In the first part of my thesis, I present IntaRNA, a new fast and accurate method for the prediction of interactions between two RNA molecules. The underlying dynamic programming algorithm incorporates the structural accessibility of interaction sites and the existence of an interaction seed region. Both features have a significant impact on the strength of RNA-RNA interactions and thus on the functionality of mRNA target sites. IntaRNA was evaluated on a dataset of experimentally verified sRNA-mRNA interactions and achieved the highest accuracy, of all compared methods, in terms of sensitivity and positive predictive value. In a genome-wide target search, IntaRNA performed as well as the best existing method, but with considerably lower computing time and memory requirement. For several sRNAs, I was able to predict whether the target is positively or negatively regulated. IntaRNA is integrated in the Freiburg RNA Tools web server to offer RNA-RNA interaction prediction for specific RNAs as well as genome-wide target searches via an easy to use web-based interface. In the second part, I present the findings of a systematic analysis of RNA-RNA interactions in which I determined features that discriminate functional from non-functional interactions and assessed the influence of these features on genome-wide target predictions. For this purpose, I compiled a set of 74 experimentally verified sRNA-target interactions and collected genome-wide full-length 5' untranslated regions. Surprisingly, I found that only interaction sites in sRNAs, but not in targets, displayed significant sequence conservation. The base pairing complementarity between sRNAs and their targets was not conserved in general across more distantly related species. In contrast to conservation, structural accessibility of functional interaction sites was significantly higher in both sRNAs and targets in comparison to non-functional sites. Based on these observations, I successfully improved the specificity of genome-wide target predictions by constraining interaction seeds to highly accessible regions in both RNAs or unstructured conserved sRNA regions. The findings on interaction site accessibility and single-stranded conserved seeds were additionally confirmed in a case study in the cyanobacterium Prochlorococcus MED4. By using an ultraconserved sequence motif of the sRNA Yfr1 as seed, two novel Yfr1 mRNA targets were predicted and subsequently experimentally confirmed. In the third part, I present PETcofold, a comparative method for the prediction of interactions and secondary structures of two multiple alignments of RNA sequences. On the premise that each of the two RNAs is structurally conserved and that conservation of their RNA-RNA interaction implies conserved function, PETcofold accounts for covariance information arising from compensatory exchanges in intra- and intermolecular base pairs. Furthermore, the method can predict pseudoknots between intra- and intermolecular base pairs by employing a hierarchical folding strategy. The ability of PETcofold to predict RNA-RNA interactions was demonstrated on a carefully curated dataset of sRNAs and their target mRNAs. On phylogenetically simulated sequences enriched for covariance patterns at the interaction sites, PETcofold performed better with increasing amounts of covariance. For evaluation of both RNA-RNA interaction and structure prediction, I exemplified that the prediction is improved by the comparative approach in comparison to existing single sequence-based methods. In summary, my thesis presents a new approach for fast and accurate RNA-RNA interaction prediction, one of the first systematic studies on accessibility and conservation of interacting sRNAs and mRNAs, the identification of two novel targets of the cyanobacterial sRNA Yfr1, and the first comparative method to predict joint secondary structures of two interacting RNAs.
Author: Richter, Andreas S.
Institution: University of Freiburg
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W6999435253
