| The ability of RNA to perform diverse cellular functions depends on its capability to form complex structures. Therefore, determining RNA structure is critical to understanding RNA function. Computational methods allow for quick determination of RNA structures, but are often prone to inaccuracies in their predictions. A newly developed technology, known as SHAPE, can be used to probe RNA structure and identify nucleotides that are likely to be single stranded and base paired. This SHAPE data can be inputted into an RNA structure program to refine predictions. Previous studies have shown that the incorporation of SHAPE data can increase the accuracy of prediction by over 30% compared to traditional mFold class algorithms. In this work, I utilize SHAPE technology to refine RNA predictions and solve new challenges. First, I create an algorithm, ShapeKnots, which incorporates SHAPE data and the prediction of pseudoknots. Pseudoknots are relatively rare RNA structural motifs that have a tendency of occurring in functional regions, but, due to their complexity, are often eliminated from structural prediction. Second, I utilize the ShapeKnots algorithm to identify pseudoknots in HIV-1 and test their role in viral replication. Third, I develop a modified partition function calculation to identify the de novo accuracy of secondary structure predictions. This allows end users to not only obtain a predicted structure, but also, to know the confidence of that prediction. Fourth, I utilize SHAPE-directed folding to identify potential alternative structures in the ribosome. Finally, I create a method to identify the accuracy of tertiary structure predictions. This allows for a quantitative measurement of accuracy when comparing predicted tertiary structures with previously determined conventional structures. |
