10.4.3 Structural Heterogeneity
Averaging over different conformational states of one particle will result in a
non-coherent average with a reduced overall resolution. Heterogeneity may be
compositional or conformational and both may be analysed by the subtomogram
classification pipelines described in Chap. 9 of this book. Compositional heterogeneity with a limited number of classes may represent bound interaction partners
and is well analysed by principal component-based statistics (Forster et al. 2008).
The multi-reference alignment (MRA) family of algorithms combines iterative
alignment of each particle to several references [78] which may efficiently sample
broad conformational spectrums of molecules. However, the resolution of each of
the classes may be limited by the number of particles contributing to it, therefore
defining the number of classes is a sensitive parameter for the final resolution. The
repeating elements from different classes may be mutually aligned and further
summed increasing the resolution [64]. Conformational heterogeneity is one of the
key challenges in both StA and in single particle cryo-EM; it should be addressed in
the following years.
10.4.4 Data Collection Strategy
Movement and dose-dependent feature degradation of cryo-preserved samples [79]
suggests a different approach to data collection rather than for getting a good point
spread function. Best data come from lowly tilted micrographs and from the first
electrons that expose the sample. It is possible to use the entire angular range for
alignment of the particles to the average and then only incorporate the first exposures corresponding to i.e. 0°–30° for the final average [64]. Pfeiffer and colleagues
recorded data starting from −20° going up to +60, then recording [−22:−2:−60] and
used only the data from −20 to 20 for the final reconstruction [60]. The low-tilt
consideration has been incorporated into an empirically derived kernel for convolution of particles and templates in Fourer space [63]. In this approach
high-resolution information is down-weighted with applied electron dose and
higher tilting angles; this was found beneficial and improved the resolution of StA.
The concept of recording the low-tilt micrographs first in the tilt series was
implemented by Hagen and colleagues as a “dose-symmetric tilt scheme” [66]. It’s
advantages are: (1) better distribution of electron dose and a more smooth effect of
beam-induced density motion on the reconstruction quality, therefore (2) ability to
apply higher electron dose per tomogram; (3) possibility to discard the information
from the highly tilted images easily and (4) higher optical stability which is
important while using phase plates. This data collection strategy was used to obtain
the highest resolution StA structure to date [67].
Finally, the angular coverage of tomograms needs to be enough to precisely align
the particles to the average but not more. Schur and colleagues used tomograms
10 Resolution in Electron Tomography
277
Averaging over different conformational states of one particle will result in a
non-coherent average with a reduced overall resolution. Heterogeneity may be
compositional or conformational and both may be analysed by the subtomogram
classification pipelines described in Chap. 9 of this book. Compositional heterogeneity with a limited number of classes may represent bound interaction partners
and is well analysed by principal component-based statistics (Forster et al. 2008).
The multi-reference alignment (MRA) family of algorithms combines iterative
alignment of each particle to several references [78] which may efficiently sample
broad conformational spectrums of molecules. However, the resolution of each of
the classes may be limited by the number of particles contributing to it, therefore
defining the number of classes is a sensitive parameter for the final resolution. The
repeating elements from different classes may be mutually aligned and further
summed increasing the resolution [64]. Conformational heterogeneity is one of the
key challenges in both StA and in single particle cryo-EM; it should be addressed in
the following years.
10.4.4 Data Collection Strategy
Movement and dose-dependent feature degradation of cryo-preserved samples [79]
suggests a different approach to data collection rather than for getting a good point
spread function. Best data come from lowly tilted micrographs and from the first
electrons that expose the sample. It is possible to use the entire angular range for
alignment of the particles to the average and then only incorporate the first exposures corresponding to i.e. 0°–30° for the final average [64]. Pfeiffer and colleagues
recorded data starting from −20° going up to +60, then recording [−22:−2:−60] and
used only the data from −20 to 20 for the final reconstruction [60]. The low-tilt
consideration has been incorporated into an empirically derived kernel for convolution of particles and templates in Fourer space [63]. In this approach
high-resolution information is down-weighted with applied electron dose and
higher tilting angles; this was found beneficial and improved the resolution of StA.
The concept of recording the low-tilt micrographs first in the tilt series was
implemented by Hagen and colleagues as a “dose-symmetric tilt scheme” [66]. It’s
advantages are: (1) better distribution of electron dose and a more smooth effect of
beam-induced density motion on the reconstruction quality, therefore (2) ability to
apply higher electron dose per tomogram; (3) possibility to discard the information
from the highly tilted images easily and (4) higher optical stability which is
important while using phase plates. This data collection strategy was used to obtain
the highest resolution StA structure to date [67].
Finally, the angular coverage of tomograms needs to be enough to precisely align
the particles to the average but not more. Schur and colleagues used tomograms
10 Resolution in Electron Tomography
277