are combined in the Fourier space to take advantage of the precise positioning of the
missing information in the Fourier transform. Thus, intensities for each frequency are
generated by averaging non-null values, in the Fourier space, from tomograms to be
combined.
7.4.1.2 Sub-Tomogram Averaging
Sub-tomogram averaging consists in performing many independent reconstructions
of biochemically identical or structurally similar objects (often macromolecular
complexes) and then to average them to get a final reconstruction having higher
resolution, less noise and without missing information artifacts [41]. To this purpose
the objects are extracted from a single or from several tomograms. Since each one of
the extracted object does not have the same orientation with respect to tilt axis, their
missing information is different. Therefore, when combined, the lack of information
of one extracted object is compensated by other objects. Such in N-axis tomography
the information is combined in the Fourier space.
The computational process for sub-tomogram averaging is mainly adapted from
single particle analysis approach [42]. This requires data alignment, classification
and reconstruction averaging. Because of the missing information occurring in
different orientation in each sub-tomogram, the major difficulty is the 3D alignment
which need to consider this lack of information to compute the correlation functions
in the Fourier space. Presently there are two main software devoted to
sub-tomogram averaging: Dynamo [43] and Relion [44].
7.4.2 Serial Images Alignment
Nowadays an important effort is realized to get 3D information by other tomographic methods which are not based on the acquisition of tilt-series. In structural
biology these methods are based on sample serial sectioning followed by image
acquisition of each section. The most widespread methods are the serial block face
(SBF) and dual-beam (FIB–SEM) in scanning electron microscopes and serial
sectioning in transmission electron microscopy (ssTEM) or array tomography in
scanning electron microscopes or light microscopy. The main differences between
the first two methods and the latest ones are based on how sections are obtained.
For SBF and FIB–SEM, resin embedded biological samples are directly sectioned
and imaged in the electron microscope whereas in the other cases sections are
produced before being deposited onto a support which is transferred to the
microscope. Once serial images acquired, from the image processing point of view,
the reconstruction process is identical independently of the acquisition method:
images are superposed to produce a 3D reconstruction.
The limitation of this technique is associated to the anisotropy in the voxel size
because X, Y dimensions depends on the electron microscope whereas Z dimension
204
A. Verguet et al.
missing information in the Fourier transform. Thus, intensities for each frequency are
generated by averaging non-null values, in the Fourier space, from tomograms to be
combined.
7.4.1.2 Sub-Tomogram Averaging
Sub-tomogram averaging consists in performing many independent reconstructions
of biochemically identical or structurally similar objects (often macromolecular
complexes) and then to average them to get a final reconstruction having higher
resolution, less noise and without missing information artifacts [41]. To this purpose
the objects are extracted from a single or from several tomograms. Since each one of
the extracted object does not have the same orientation with respect to tilt axis, their
missing information is different. Therefore, when combined, the lack of information
of one extracted object is compensated by other objects. Such in N-axis tomography
the information is combined in the Fourier space.
The computational process for sub-tomogram averaging is mainly adapted from
single particle analysis approach [42]. This requires data alignment, classification
and reconstruction averaging. Because of the missing information occurring in
different orientation in each sub-tomogram, the major difficulty is the 3D alignment
which need to consider this lack of information to compute the correlation functions
in the Fourier space. Presently there are two main software devoted to
sub-tomogram averaging: Dynamo [43] and Relion [44].
7.4.2 Serial Images Alignment
Nowadays an important effort is realized to get 3D information by other tomographic methods which are not based on the acquisition of tilt-series. In structural
biology these methods are based on sample serial sectioning followed by image
acquisition of each section. The most widespread methods are the serial block face
(SBF) and dual-beam (FIB–SEM) in scanning electron microscopes and serial
sectioning in transmission electron microscopy (ssTEM) or array tomography in
scanning electron microscopes or light microscopy. The main differences between
the first two methods and the latest ones are based on how sections are obtained.
For SBF and FIB–SEM, resin embedded biological samples are directly sectioned
and imaged in the electron microscope whereas in the other cases sections are
produced before being deposited onto a support which is transferred to the
microscope. Once serial images acquired, from the image processing point of view,
the reconstruction process is identical independently of the acquisition method:
images are superposed to produce a 3D reconstruction.
The limitation of this technique is associated to the anisotropy in the voxel size
because X, Y dimensions depends on the electron microscope whereas Z dimension
204
A. Verguet et al.