J.M.Carazo, Paul Eggermont (*), G. Herman (+) and S.Scheres
Centro Nacional de Biotecnología, CSIC, Cantoblanco, 28049, Madrid, Spain
(+)The Graduate Center, City University of New York, New York, NY10016, USA
(*) Food and Resource Economics, University of Delaware, Newark, DE19716, USA
Back to Image Analysis Challenges in Molecular Microscopy
Macromolecular manomachines refer to those complexes that correspond to the functional machinery in charge of defined but varied functions in the context of the living cell, from DNA repair or replication, to protein degradation, to name just a few functions. Normally, they are formed by different types of proteins and may also include nucleic acids. Their weight ranges typically from a few hundreds of kilo Daltons, to the million Dalton, with a size in the range of the tens of nanometer. Their study is normally accomplished by three-dimensional electron microscopy, since they are too large for Nuclear Magtenic Resonance and perhaps too flexible for X-ray Diffraction. Indeed, we will focus our attention to their conformational flexibility, since it is very common that the way to carry on their function is by alternating between different conformations.
The way three-dimensional electron microscopy obtains structural information from these macromolecular nanomachines is conceptually similar to the way a CT scanner works in medical imaging. However, there is a very important experimental difference in the fact that we cannot control the relative position of each of these nanomachines with respect to the electron beam, so that the three-dimensional reconstruction problem is the one of reconstructing the “average” structure from thousands of very noisy projection images, with the “complication” that the orientation for each nanomachine has to be estimated too. Furthermore, in this context flexibility appears as an additional variable in the reconstruction process. The situation is not now the reconstruction of ONE average structure, but we have to classify each projection image by its projection geometry AND also by its “structural class”.
We will present our work so far in casting the three-dimensional electron microscopy reconstruction problem in the context of a Maximum Likelihood approach from which we derive the projection geometry for each image and the structural class and then perform a series of 3d-reconstructions.