The analysis of the relations between structure and function of a biopolymer such as RNA is a core issue in structural biology and bioinformatics. The folding process of RNA structure follows a strict hierarchical scheme. The native structure of an RNA molecule is formed by the subsequent formation of secondary and tertiary structure motifs. These motifs are common in RNA molecules of different species, suggesting that parts (fragments) of a molecule with known structure can be used to predict unknown RNA structures. This thesis describes methods for an approach on the modeling of RNA three-dimensional structure based on structure fragments. We describe the design and composition of a novel template fragment library, containing about 25,000 template fragments from 578 different source structures. Based on this library, we semi-automatically model selected RNA structures, achieving 70% coverage of the native structure of an RNAse P A-type S domain. In addition, we present implementational details for a semi-automatic modeling framework employing an RNA sequence-structure alignment and an alignment validation method via isostericity matrices. The framework is tested on sample hairpin and tRNA targets. The results of this thesis demonstrate that building a target structure based on structure fragments from different template structures is a feasible approach for the RNA structure modeling and prediction problem.
The prediction of RNA structure is a major field of bioinformatics research. A framework for modeling RNA 3D structures based on sequence-to-structure alignment via genetic algorithm (RAGA) has been designed in a previous work. Due to performance and robustness problems with the RAGA algorithm it has to be exchanged with a newly developed method: Lagrangian Relaxed RNA structure alignment (LR-RSA). This thesis describes the necessary modifications which had to be applied to the framework to embed the new alignment and restore the operative status of the RNA Threading Framework. Also a validation is given whether the new algorithm is an acceptable replacement within the RNA 3D structure prediction.