4.4 Data Processing
for High-Resolution
Tomography
Data processing for high-resolution tomography mirrors that of
conventional data processing, but with additional steps to optimize
data. Recently, user-friendly software packages have been developed (EMAN2, emClarity, relion); however, to date every novel
high-resolution subtomogram averaging structure has been based
on a workflow utilizing multiple software packages and personal
scripts [27, 28, 38, 42].
1. Frame alignment: gain-correct, rotate, and align the tilt frames
in the software of your choice. IMOD alignframes or MotionCor2 both perform well for frame alignment.
2. Dose-weighting: tilts should be dose-weighted prior to tilt
stack alignment. All available software methods are based on
the single-particle averaging estimation of the relationship
between electron dose and final resolution from the Grigorieff
group [45].
3. CTF estimation: the CTF can be estimated with any standard
software package. However, CTF estimation software routinely
underperforms at high tilt angles. It is common to test multiple
packages for different data collections or to use homemade
software to assess or improve outputs.
4. Reconstruction: the most common reconstruction method is
weighted
back
projection,
coupled
with
either
two-dimensional CTF correction in IMOD or threedimensional CTF correction in NovaCTF [46]. For
low-defocus tomograms, an SIRT-like filter can be applied
during reconstruction for visualization purposes.
5. Subtomogram averaging: for a low particle count (100s), high
resolution is not expected and PEET or RELION may be used
as in conventional tomography. For particle counts above
10,000, high-resolution subtomogram averaging often
involves parallelized GPU processing and a highly samplespecific workflow. Dynamo and AV3 have both performed to
high resolution on novel structures [28, 42] and are flexible for
a customized workflow. emClarity and EMAN2 have performed to high resolution on a dataset previously solved in
AV3, but are less flexible for advanced users.
4.5 Subtomogram
Averaging
In subtomogram averaging, certain subvolumes are cropped from
the tomogram, aligned, and then averaged to enhance the signalto-noise ratio of a certain repetitive feature (e.g., a protein complex
of interest) within a single tomogram or across different tomograms of the same species collected with similar pixel size. Many
software packages are currently available to perform subtomogram
averaging, including: Dynamo, PEET, RELION, EMAN2, and
PyTOM (see, e.g., Refs. 47–52).
106
Mohammed Kaplan et al.
for High-Resolution
Tomography
Data processing for high-resolution tomography mirrors that of
conventional data processing, but with additional steps to optimize
data. Recently, user-friendly software packages have been developed (EMAN2, emClarity, relion); however, to date every novel
high-resolution subtomogram averaging structure has been based
on a workflow utilizing multiple software packages and personal
scripts [27, 28, 38, 42].
1. Frame alignment: gain-correct, rotate, and align the tilt frames
in the software of your choice. IMOD alignframes or MotionCor2 both perform well for frame alignment.
2. Dose-weighting: tilts should be dose-weighted prior to tilt
stack alignment. All available software methods are based on
the single-particle averaging estimation of the relationship
between electron dose and final resolution from the Grigorieff
group [45].
3. CTF estimation: the CTF can be estimated with any standard
software package. However, CTF estimation software routinely
underperforms at high tilt angles. It is common to test multiple
packages for different data collections or to use homemade
software to assess or improve outputs.
4. Reconstruction: the most common reconstruction method is
weighted
back
projection,
coupled
with
either
two-dimensional CTF correction in IMOD or threedimensional CTF correction in NovaCTF [46]. For
low-defocus tomograms, an SIRT-like filter can be applied
during reconstruction for visualization purposes.
5. Subtomogram averaging: for a low particle count (100s), high
resolution is not expected and PEET or RELION may be used
as in conventional tomography. For particle counts above
10,000, high-resolution subtomogram averaging often
involves parallelized GPU processing and a highly samplespecific workflow. Dynamo and AV3 have both performed to
high resolution on novel structures [28, 42] and are flexible for
a customized workflow. emClarity and EMAN2 have performed to high resolution on a dataset previously solved in
AV3, but are less flexible for advanced users.
4.5 Subtomogram
Averaging
In subtomogram averaging, certain subvolumes are cropped from
the tomogram, aligned, and then averaged to enhance the signalto-noise ratio of a certain repetitive feature (e.g., a protein complex
of interest) within a single tomogram or across different tomograms of the same species collected with similar pixel size. Many
software packages are currently available to perform subtomogram
averaging, including: Dynamo, PEET, RELION, EMAN2, and
PyTOM (see, e.g., Refs. 47–52).
106
Mohammed Kaplan et al.