Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
Initialising ...
松本 淳; 岩崎 憲治*; 高木 淳一*
no journal, ,
We have developed a computational approach to build an atomic model from an electron microscope (EM) image of proteins. In this approach, many atomic models are built first by deforming the X-ray crystal structure of the protein using a computational technique. Then, projection images of each atomic model to many different directions are generated. Finally, they are compared to the EM image. The atomic models with the projection images similar to the EM image are regarded as the candidates for the atomic structure of the protein from which the EM image was obtained. This approach was applied to the EM images of integrin. The X-ray crystal structure of integrin was solved for the closed conformation, while many EM images were obtained for the open conformation. The difference of the two conformations is so large that it is difficult to build the atomic model for the EM conformations by deforming the X-ray crystal structure using conventional computational techniques. Thus, we have used the coarse-grained model of the X-ray crystal structure to simulate the large-scale conformational change.
岩崎 憲治*; 松本 淳; 高木 淳一*
no journal, ,
We developed a method for comparing 2-D electron microscopy (EM) images to a pool of deformed atomic structures that reduces the susceptibility of subjective interpretation and is more expeditious than 3-D analysis. Firstly, we used normal mode analysis (NMA) in the context of an elastic network model (ENM) to prepare a pool of deformed structures from coarse-grain (CG) structural models obtained using Ca atom coordinates from crystal structures. Then, small deformations were applied to the model along low frequency normal modes. Both NMA and deformation were performed iteratively to build many different deformed atomic structures. 2-D images were then compared to these structures to objectively locate their model of best fit. The fitting-search is performed as a rigid body fitting, which works effectively for very noisy raw images as well as for molecules with a large conformational change. Our most representative example of this hybrid approach of using 2-D EM images and crystal structures has been the study of the integrin cell-adhesion molecules. Our hybrid method has allowed the investigation of the conformationally flexible integrins, and is providing useful evidence of whether other members of that family also use the switch-blade model observed in integrin avb3.