Masterprüfung mit Defensio, Thomas Spicher

- 30.11.2021 15:30

„Computational Estimation of Energy Parameters for RNA folding“

The structure of RNA is generally crucial for its biological function. Most of the structural information needed to understand RNA functioning can be gathered from its secondary structure. RNA secondary structure prediction relies on energy parameters corresponding to the free energies of small structure motifs. For canonical nucleotides, these parameters have been derived from UV melting experiments and collected in the Nearest Neighbor (NN) parameter database. NN parameters are used in most of the physics based algorithms for secondary structure prediction. Although the Watson-Crick base pairs represent the most important motifs in RNA secondary structures, the lack of parameters for modified nucleotides impede the computational prediction of RNA structures. Only a few parameters for modified nucleotides have been determined and the effort to experimentally measure all relevant parameters is unrealistically high. An alternative to expensive experiments is to extract parameters from computational simulations of small structure motifs. The Rosetta software provides a coarse grained model, the RECCES framework, which can realize these kinds of simulations. In previous work, the Rosetta-RECCES framework was shown to be capable of predictions of NN parameters containing inosine, which are comparable to experimental results. In order to complete the existing NN parameters and in further steps to improve the RNA secondary structure prediction algorithms, the aim of this work is to examine the ability of computational simulations to determine NN parameters of stacking pairs containing modified nucleotides. Two different methods were used to computationally extract energy parameters: subtraction of duplexes differing by exactly one stack and linear regression. In both cases the computational predictions for modified nucleotides lie outside the range of the experimentally determined NN parameter values. While the analysis of the folding free energy predicted by RECCES shows high correlation with estimations from the NN model and existing energy parameters, a modification dependent systematic error was found. This may indicate some imprecision either of the Rosetta energy function or of the simulation behaviour when used on modified nucleotides. Despite the exposed limitations of the RECCES framework concerning I • U, Ψ • A, and m6A • U base pairs, the accurate predictions for Watson-Crick, G • U, and I • C base pairs encourage to continue the research on coarse grained models to predict new NN parameters.



online Videokonferenz