Tomographic reconstruction of deformable objects for electron cryo microscopy – RHODES

The study of macromolecule functions in the cell is very usefull to understand cellular life and some viral diseases like AIDS. These functions are linked to the macromolecule 3D structure and to its deformations. Theses deformations correspond to a sequel of states, so called conformations, of the macromolecule. Then, the determination of the macromolecule deformations is a key point for the progress of medicine.

The Cryo electron Microscopy (CM) is one the the main modalities to study macromolecules. It enables, after reconstruction, to visualize the 3D volume of large macromolecules (1 000-10 000 atomes) with a resolution above the nanometer.
The CM data consist of a large number (10 000-1 000 000) of images, each including a tomographic projection of a single specimen of the same macromolecule. For the images, the orientation and the state of deformation (corresponding to the observed specimen) are unknown. Therefore, reconstruction of a 3D object from images obtained by cryo-electron microscopy is a difficult problem with important biological issues. The resolution of the the macromolecule 3D volumes reconstructed by CM is about a nanometer.

Currently, reconstruction methods are based on the classification of acquired images in a finite number of distinct classes corresponding to the different orientations of the observed object, which are then subdivided in subclasses corresponding to the different conformations. However, these methods are not suitable to the case where the macromolecule deformation is continuous. To handle this case, we propose in the project “tomogRapHic reconstruction for electron cryO-microscopy DEformable objectS” so called RHODES to model the all the 3D volumes of the macromolecule as a continuous set V. The dimension of V is the sum of the three spatial dimensions and the dimensions corresponding to the deformation parameters.

Two leads are proposed to reconstruct the continuous set V: a) a dimension reduction for the parameter estimation followed by the reconstruction of V based on these parameters ; b) a method based on the minimization of a cost evaluating the reconstruction parameters and the reconstructed volume.
Several steps are necessary for this study, besides the implementation of two methods of reconstruction: the development of artificial data to control noise and deformations, the acquisition of real data, tests and validations on artificial and real data.

The algorithms and data produced during the RHODES project will be used to rebuild three macromolecules exhibiting continuous deformations.
First the ribosome (that is well studied and will be used to validate the developped methods), and two macromolecules that have a strong biological interest but for wich the continuous deformation is the main limitation to the spatial resolution of 3D volume:
- human immunodeficiency virus integrase (the virus causing AIDS),
- the transcription factor TFIID.

Among the nine researchers involved in the project RHODES, six are computer scientists expert in image processing. They will develop reconstruction methods adapted to the case of spatial continuous deformation by applying the methods described above. The three other researchers are experts in structural biology. They will participate in the development and validation of the developed techniques and will acquire real data necessary to the volume reconstruction of the three chosen macromolecules.