Rigidity analysis of protein structures and rapid simulations of protein motion

摘要

It is a common goal in biophysics to understand protein structural properties and their relationship to protein function. I investigated protein structural properties using three coarse graining methods: a rigidity analysis method First, a geometric simulation method Froda and normal mode analysis as implemented in Elnemo to identify the protein directions of motion. Furthermore, I also compared the results between the coarse graining methods with the results from molecular dynamics and from experiments that I carried out. The results from the rigidity analysis across a set of protein families presented in chapter 3 highlighted two different patterns of protein rigidity loss, i.e. "sudden" and "gradual". It was found that theses characteristic patterns were in line with the rigidity distribution of glassy networks. The simulations of protein motion by merging flexibility, rigidity and normal mode analyses presented in chapter 4 were able to identify large conformational changes of proteins using minimal computational resources. I investigated the use of RMSD as a measure to characterise protein motion and showed that, despite it is a good measure to identify structural differences when comparing the same protein, the use of extensive RMSD better captures the extend of motion of a protein structure. The in-depth investigation of yeast PDI mobility presented in chapter 5 confirmed former experimental results that predicted a large conformational change for this enzyme. Furthermore, the results predicted: a characteristic rigidity distribution for yeast PDI, a minimum and a maximum active site distance and a relationship between the energy cutoff, i.e. the number of hydrogen bonds part of the network of bonds, and protein mobility. The results obtained were tested against molecular dynamics simulations in chapter 6. The MD simulation also showed a large conformational change for yeast PDI but with a slightly different minimum and maximum inter-cysteine distance. Furthermore, MD was able to reveal new data, i.e. the most likely inter-cysteine distance. In order to test the accuracy of the coarse graining and MD simulations I carried out cross-linking experiments to test the minimum inter-cysteine distance predictions. The results presented in chapter 7 show that human PDI minimum distance is below 12Å whereas the yeast PDI minimum distance must be above 12Å as no cross-linking structures where found with the available (12Å long) cross-linkers.