Mapping Protein Unfolding using Coulomb Explosions : A theoretical study on unfolding using Molecular Dynamics

Sammanfattning

The field of structure determination using high intensity lasers enables diffraction patterns of single molecular systems to be obtained without the need of crystallization. This defines Single-Particle Imaging (SPI). With ongoing development in the area, molecular dynamics (MD) grows more popular and accurate, being able to simulate particle movement under specific conditions. One side effect of SPI is strong ionizations, resulting in a violent explosion of the system, called a Coulomb explosion. However, what if there were information to be retrieved from the ion trajectories of the explosion? Based on the widely used MD engine GROMACS, a toolbox for simulating ionizations of atoms under high-intensity laser conditions has been developed, called MolDStruct. Using this toolbox, ion trajectories from ubiquitin, well-studied protein, has been produced and detected on a simulated, plane detector. The aim of this thesis is to show if there is a way to track the structural changes of ubiquitin at different levels of unfolding only using ion maps. It involves examining if unfolding pathways can be established and if there is a way to define the degree of unfolding using only ion maps. 101 structures spanning over a 50 ps long electric field simulation on ubiquitin were extracted and fed into MolDStruct. The ion trajectory output data has been mapped on a flat plane detector at two different axes in space in order to study the explosion patterns. This was done in depth for two single datasets of electric field simulations at one electric field strength and also on two large datasets consisting of 100 separate electric field simulations at two different electric field strengths. The analysis of the ion maps went in two directions. One was to apply dimensionality reduction techniques PCA, t-SNE and UMAP to explore the clustering capabilities of the ion map data. The other direction was to find correlations between the output and the actual unfolding of ubiquitin. Results from dimensionality reduction methods show great clustering capabilities for single datasets but difficulties when using all datasets. Folded and unfolded cases can, however, be separated by these techniques. Additionally, linear correlations could be found relating to ubiquitin’s unfolding. The first correlation was found by plotting the evolution of the ion-spread on the detector as a measure of x- and y-diameters of an ellipse to the evolution of protein length calculated from radius of gyration. Secondly, another correlation could be found when the number of ions hitting the detector was plotted against ubiquitin’s radius of gyration. This shows that there are ways to track the level of unfolding using the information given by ion detectors. By understanding the capabilities of ion detection, in tandem with applying above mentioned methods, it can refine and facilitate SPI as we know it today.