Reproducibility of protein explosions with an X-ray laser

Sammanfattning

Protein structures are important to determine in order to understand their function. Today, there are many methods used to structurally determine molecules, one of the most popular being X-ray crystallography. X-ray crystallography presents many positive aspects, including an amplified signal, along with protection of the sample against X-ray radiation damage. The biggest challenge for X-ray crystallography has to do with the simple fact that not all proteins are able to be crystallized. Single Particle Imaging (SPI) is an alternative method, which simply involves imaging one protein at a time, but comes at the cost of the structure quickly decaying due to the high-intensity X-rays on the sample. Much research has therefore been focused on developing ultra-fast lasers to image the protein before it is destroyed. This study aims to evaluate how the protein destruction may instead yield structural information about the protein, after its decay. This project investigates the ability to detect small key structural differences in proteins of high similarity, using information from the radiation damage of the proteins. In this study, multiple different experimental parameters, such as X-ray dynamics and detector geometry, were varied to see their effect on the distinguishability between proteins. Five different ubiquitin proteins were analyzed, four being bound to chromophores at different locations. This was done in order to have highly similar, albeit different structures. The aim of this study is to guide future research in the structure determination field and to find the parameters that are important for distinguishing between structures using ion data from Coulomb explosions. All the studied protein structures could be distinguished to a high degree when analyzing the different ubiquitins at fixed orientations. However, as soon as differences in molecular rotations are introduced, distinguishability using the ion data from the Coulomb explosions becomes difficult. The results suggest that the choice of pulse- and detector parameters are important, with better distinguishability occurring with detectors close to the explosion with a low number of bins. The pulse should ideally cause high amounts of fast ionizations, which depends on the X-ray photon energy used. Out of the studied photon energies, 300 eV and 600 eV resulted in much better distinguishability than when using photon energies of 2000 eV. An equally important factor is the choice of classification algorithm used to analyze the results, and the data representation.