Usage
Features in a tomogram that resemble a structural 'template' can be localized in an automated fashion using 'template matching'. In this approach a 3D template is correlated with a given tomogram. In this procedure the different possible rotations and translations are sampled exhaustively using the algorithm described in Förster et al. (2010).
Requirements
For usage you need at least a set of reconstructed tomograms in the MRC format and a template structure in the MRC format. Tomograms in IMOD format (.rec) are also okay but need to be renamed (or softlinked!) to have the correct extension (.mrc). Tomograms are ideally binned 6x or 8x to prevent excessive runtimes. The template can be an EM reconstruction (from the EMDB) or a PDB that was coverted to a density (for example via Chimera molmap).
Template matching workflow
Using template matching in this software consists of the following steps:
- Creating a template and mask
- Matching the template in a tomogram
- Extracting particles
- Merging annotations for export to other software
1. Creating a template and mask
Keep in mind:
- The template and mask need to have the same box size.
- The template needs to have the same contrast as the tomogram (e.g. the particles
are black in both the tomogram and template). Contrast can be adjusted with the
--invertoption.
pytom_create_template.py
Using an EM map as a reference structure generally leads to the best results. Alternatively a structure from the PDB can be converted in Chimera(X) using the molmap command to create an MRC file that models the electrostatic potential. A good ballpark for the box size of the template is 2 or 3 times the particle diameter (along its longest axis).
usage: pytom_create_template.py [-h] -i INPUT_MAP [-o OUTPUT_FILE]
[--input-voxel-size-angstrom INPUT_VOXEL_SIZE_ANGSTROM]
--output-voxel-size-angstrom
OUTPUT_VOXEL_SIZE_ANGSTROM [--center CENTER]
[--low-pass LOW_PASS] [-b BOX_SIZE]
[--invert INVERT] [-m MIRROR] [--log LOG]
Generate template from MRC density. -- Marten Chaillet (@McHaillet)
options:
-h, --help show this help message and exit
-i INPUT_MAP, --input-map INPUT_MAP
Map to generate template from; MRC file.
-o OUTPUT_FILE, --output-file OUTPUT_FILE
Provide path to write output, needs to end in .mrc .
If not provided file is written to current directory
in the following format:
template_{input_map.stem}_{voxel_size}A.mrc
--input-voxel-size-angstrom INPUT_VOXEL_SIZE_ANGSTROM
Voxel size of input map, in Angstrom. If not provided
will be read from MRC input (so make sure it is
annotated correctly!).
--output-voxel-size-angstrom OUTPUT_VOXEL_SIZE_ANGSTROM
Output voxel size of the template, in Angstrom. Needs
to be equal to the voxel size of the tomograms for
template matching. Input map will be downsampled to
this spacing.
--center CENTER Set this flag to automatically center the density in
the volume by measuring the center of mass.
--low-pass LOW_PASS Apply a low pass filter to this resolution, in
Angstrom. By default a low pass filter is applied to a
resolution of (2 * output_spacing_angstrom) before
downsampling the input volume.
-b BOX_SIZE, --box-size BOX_SIZE
Specify a desired size for the output box of the
template. Only works if it is larger than the
downsampled box size of the input.
--invert INVERT Multiply template by -1. WARNING: not needed if ctf
with defocus is already applied!
-m MIRROR, --mirror MIRROR
Mirror the final template before writing to disk.
--log LOG Can be set to `info` or `debug`
pytom_create_mask.py
The mask around the template can be quite tight to remove as much noise as possible around the particles of interest. We recommend around 10%-20% overhang relative to the particle radius. You can also generate an ellipsoidal mask for particles that do not approximate well as a sphere. Though you will probably need to reorient this mask in chimera and resample to the grid of the template. Optionally you could also create a structured mask around the template in external software (via thresholding and dilation for example). Take into account that non-spherical masks roughly double the template matching computation time.
usage: pytom_create_mask.py [-h] -b BOX_SIZE [-o OUTPUT_FILE]
[--voxel-size VOXEL_SIZE] -r RADIUS
[--radius-minor1 RADIUS_MINOR1]
[--radius-minor2 RADIUS_MINOR2] [-s SIGMA]
Create a mask for template matching. -- Marten Chaillet (@McHaillet)
options:
-h, --help show this help message and exit
-b BOX_SIZE, --box-size BOX_SIZE
Shape of square box for the mask.
-o OUTPUT_FILE, --output-file OUTPUT_FILE
Provide path to write output, needs to end in .mrc .If
not provided file is written to current directory in
the following format:
./mask_b[box_size]px_r[radius]px.mrc
--voxel-size VOXEL_SIZE
Provide a voxel size to annotate the MRC (currently
not used for any mask calculation).
-r RADIUS, --radius RADIUS
Radius of the spherical mask in number of pixels. In
case minor1 and minor2 are provided, this will be the
radius of the ellipsoidal mask along the x-axis.
--radius-minor1 RADIUS_MINOR1
Radius of the ellipsoidal mask along the y-axis in
number of pixels.
--radius-minor2 RADIUS_MINOR2
Radius of the ellipsoidal mask along the z-axis in
number of pixels.
-s SIGMA, --sigma SIGMA
Sigma of gaussian drop-off around the mask edges in
number of pixels. Values in the range from 0.5-1.0 are
usually sufficient for tomograms with 20A-10A voxel
sizes.
2. Matching the template in a tomogram
pytom_match_template.py
This script requires at least a tomogram, a template, a mask, the min and max tilt angles (for missing wedge constraint), an angular search, and a GPU index to run. The search can be limited along any dimension with the --search-x, --search-y, and --search-z parameters; for example to skip some empty regions in the z-dimension where the ice layer is not yet present, or to remove some reconstruction artifact region along the x-dimension. With the --volume-split option, a tomogram can be split into chunks to allow them to fit in GPU memory (useful for large tomograms). Providing multiple GPU's will allow the program to split the angular search (or the subvolume search) over multiple cards to speed up the algorithm.
The software automatically calculates the angular search based on the available resolution and provided particle diameter. The required search is found from the Crowther criterion \(\Delta \alpha = \frac{180}{\pi r_{max} d}\). For the maximal resolution the voxel size is used, unless a low-pass filter is specified as this limits the available maximal resolution. You can exploit this to reduce the angular search! For non-spherical particles we suggest choosing the particle diameter as the longest axis of the macromolecule.
In case the template matching is run with a non-spherical mask, it is essential to set the --non-spherical-mask flag. It requires a slight modification of the calculation that will roughly double the computation time, so only use non-spherical masks if absolutely necessary.
Optimizing results: per tilt weighting with CTFs and dose accumulation
Optimal results are obtained by also incorporating information for the 3D CTF. You can pass the following files (and parameters):
- Tilt angles: a
.rawtltor.tltfile to the--tilt-anglesparameter with all the tilt angles used to reconstruct the tomogram. You should then also set the--per-tilt-weightingflag. - CTF data: a
.defocusfile from IMOD or.txtfile to--defocus-file. The. txtfile should specify the defocus of each tilt in \(\mu m\). You can also give a single defocus value (in \(\mu m\)). The CTF will also require input for--voltage,--amplitude-contrast, and--spherical-abberation. - Dose weighting: a
.txtfile to--dose-accumulationwith the accumulated dose per tilt (assuming the same ordering as.tlt). Each line contains a single float specifying the accumulated dose in \(e^{-}/\text{Å}^{2}\). Dose weighting only works in combination with--per-tilt-weighting.
(As a side note, you can also only enable --per-tilt-weighting without dose accumulation and CTFs, or with either dose accumulation or CTFs.)
When enabling the CTF model here (with the defocus file), it is important that the template is not multiplied with a CTF before passing it to this script. The template only needs to be scaled to the correct pixel size and the contrast should be adjusted to match the contrast in the tomograms.
Secondly, if the tomogram was CTF corrected, for example by using IMODs
strip-based CTF correction or NovaCTF. Its important to add the parameter
--tomogram-ctf-model phase-flip which modifies the template CTF to match the
tomograms CTF correction.
Background corrections
The software contains two background correction methods that might improve results:
--spectral-whitening or --random-phase-correction (from STOPGAP). In our
experience the random phase correction is most reliable, while spectral whitening
never seemed to clearly improve results.
usage: pytom_match_template.py [-h] -t TEMPLATE -v TOMOGRAM [-d DESTINATION]
-m MASK
[--non-spherical-mask NON_SPHERICAL_MASK]
[--particle-diameter PARTICLE_DIAMETER]
[--angular-search ANGULAR_SEARCH]
[--z-axis-rotational-symmetry Z_AXIS_ROTATIONAL_SYMMETRY]
[-s VOLUME_SPLIT VOLUME_SPLIT VOLUME_SPLIT]
[--search-x SEARCH_X SEARCH_X]
[--search-y SEARCH_Y SEARCH_Y]
[--search-z SEARCH_Z SEARCH_Z]
[--tomogram-mask TOMOGRAM_MASK]
[-a TILT_ANGLES [TILT_ANGLES ...]]
[--per-tilt-weighting PER_TILT_WEIGHTING]
[--voxel-size-angstrom VOXEL_SIZE_ANGSTROM]
[--low-pass LOW_PASS] [--high-pass HIGH_PASS]
[--dose-accumulation DOSE_ACCUMULATION]
[--defocus DEFOCUS]
[--amplitude-contrast AMPLITUDE_CONTRAST]
[--spherical-aberration SPHERICAL_ABERRATION]
[--voltage VOLTAGE] [--phase-shift PHASE_SHIFT]
[--tomogram-ctf-model {phase-flip}]
[--defocus-handedness {-1,0,1}]
[--spectral-whitening SPECTRAL_WHITENING]
[-r RANDOM_PHASE_CORRECTION]
[--half-precision HALF_PRECISION]
[--rng-seed RNG_SEED]
[--relion5-tomograms-star RELION5_TOMOGRAMS_STAR]
-g GPU_IDS [GPU_IDS ...] [--log LOG]
Run template matching. -- Marten Chaillet (@McHaillet)
options:
-h, --help show this help message and exit
Template, search volume, and output:
-t TEMPLATE, --template TEMPLATE
Template; MRC file. Object should match the contrast
of the tomogram: if the tomogram has black ribosomes,
the reference should be black.
(pytom_create_template.py has an option to invert
contrast)
-v TOMOGRAM, --tomogram TOMOGRAM
Tomographic volume; MRC file.
-d DESTINATION, --destination DESTINATION
Folder to store the files produced by template
matching.
Mask:
-m MASK, --mask MASK Mask with same box size as template; MRC file.
--non-spherical-mask NON_SPHERICAL_MASK
Flag to set when the mask is not spherical. It adds
the required computations for non-spherical masks and
roughly doubles computation time.
Angular search:
--particle-diameter PARTICLE_DIAMETER
Provide a particle diameter (in Angstrom) to
automatically determine the angular sampling using the
Crowther criterion. For the max resolution, (2 * pixel
size) is used unless a low-pass filter is specified,
in which case the low-pass resolution is used. For
non-globular macromolecules choose the diameter along
the longest axis.
--angular-search ANGULAR_SEARCH
This option overrides the angular search calculation
from the particle diameter. If given a float it will
generate an angle list with healpix for Z1 and X1 and
linear search for Z2. The provided angle will be used
as the maximum for the linear search and for the mean
angle difference from healpix.Alternatively, a .txt
file can be provided with three Euler angles (in
radians) per line that define the angular search.
Angle format is ZXZ anti-clockwise (see: https://www.c
cpem.ac.uk/user_help/rotation_conventions.php).
--z-axis-rotational-symmetry Z_AXIS_ROTATIONAL_SYMMETRY
Integer value indicating the rotational symmetry of
the template around the z-axis. The length of the
rotation search will be shortened through division by
this value. Only works for template symmetry around
the z-axis.
Volume control:
-s VOLUME_SPLIT VOLUME_SPLIT VOLUME_SPLIT, --volume-split VOLUME_SPLIT VOLUME_SPLIT VOLUME_SPLIT
Split the volume into smaller parts for the search,
can be relevant if the volume does not fit into GPU
memory. Format is x y z, e.g. --volume-split 1 2 1
--search-x SEARCH_X SEARCH_X
Start and end indices of the search along the x-axis,
e.g. --search-x 10 490
--search-y SEARCH_Y SEARCH_Y
Start and end indices of the search along the y-axis,
e.g. --search-x 10 490
--search-z SEARCH_Z SEARCH_Z
Start and end indices of the search along the z-axis,
e.g. --search-x 30 230
--tomogram-mask TOMOGRAM_MASK
Here you can provide a mask for matching with
dimensions (in pixels) equal to the tomogram. If a
subvolume only has values <= 0 for this mask it will
be skipped.
Filter control:
-a TILT_ANGLES [TILT_ANGLES ...], --tilt-angles TILT_ANGLES [TILT_ANGLES ...]
Tilt angles of the tilt-series, either the minimum and
maximum values of the tilts (e.g. --tilt-angles -59.1
60.1) or a .rawtlt/.tlt file with all the angles (e.g.
--tilt-angles tomo101.rawtlt). In case all the tilt
angles are provided a more elaborate Fourier space
constraint can be used
--per-tilt-weighting PER_TILT_WEIGHTING
Flag to activate per-tilt-weighting, only makes sense
if a file with all tilt angles have been provided. In
case not set, while a tilt angle file is provided, the
minimum and maximum tilt angle are used to create a
binary wedge. The base functionality creates a fanned
wedge where each tilt is weighted by cos(tilt_angle).
If dose accumulation and CTF parameters are provided
these will all be incorporated in the tilt-weighting.
--voxel-size-angstrom VOXEL_SIZE_ANGSTROM
Voxel spacing of tomogram/template in angstrom, if not
provided will try to read from the MRC files. Argument
is important for band-pass filtering!
--low-pass LOW_PASS Apply a low-pass filter to the tomogram and template.
Generally desired if the template was already filtered
to a certain resolution. Value is the resolution in A.
--high-pass HIGH_PASS
Apply a high-pass filter to the tomogram and template
to reduce correlation with large low frequency
variations. Value is a resolution in A, e.g. 500 could
be appropriate as the CTF is often incorrectly
modelled up to 50nm.
--dose-accumulation DOSE_ACCUMULATION
Here you can provide a file that contains the
accumulated dose at each tilt angle, assuming the same
ordering of tilts as the tilt angle file. Format
should be a .txt file with on each line a dose value
in e-/A2.
--defocus DEFOCUS Here you can provide an IMOD defocus (.defocus) file
(version 2 or 3) , a text (.txt) file with a single
defocus value per line (in μm), or a single defocus
value (in μm). The value(s), together with the other
ctf parameters (amplitude contrast, voltage, spherical
abberation), will be used to create a 3D CTF weighting
function. IMPORTANT: if you provide this, the input
template should not be modulated with a CTF
beforehand. If it is a reconstruction it should
ideally be Wiener filtered.
--amplitude-contrast AMPLITUDE_CONTRAST
Amplitude contrast fraction for CTF.
--spherical-aberration SPHERICAL_ABERRATION
Spherical aberration for CTF in mm.
--voltage VOLTAGE Voltage for CTF in keV.
--phase-shift PHASE_SHIFT
Phase shift (in degrees) for the CTF to model phase
plates.
--tomogram-ctf-model {phase-flip}
Optionally, you can specify if and how the CTF was
corrected during reconstruction of the input tomogram.
This allows match-pick to match the weighting of the
template to the tomogram. Not using this option is
appropriate if the CTF was left uncorrected in the
tomogram. Option 'phase-flip' : appropriate for IMOD's
strip-based phase flipping or reconstructions
generated with novaCTF/3dctf.
--defocus-handedness {-1,0,1}
Specify the defocus handedness for defocus gradient
correction of the CTF in each subvolumes. The more
subvolumes in x and z, the finer the defocus gradient
will be corrected, at the cost of increased computing
time. It will only have effect for very clean and
high-resolution data, such as isolated macromolecules.
IMPORTANT: only works in combination with --volume-
split ! A value of 0 means no defocus gradient
correction (default), 1 means correction assuming
correct handedness (as specified in Pyle and Zianetti
(2021)), -1 means the handedness will be inverted. If
uncertain better to leave off as an inverted
correction might hamper results.
--spectral-whitening SPECTRAL_WHITENING
Calculate a whitening filtering from the power
spectrum of the tomogram; apply it to the tomogram
patch and template. Effectively puts more weight on
high resolution features and sharpens the correlation
peaks.
Additional options:
-r RANDOM_PHASE_CORRECTION, --random-phase-correction RANDOM_PHASE_CORRECTION
Run template matching simultaneously with a phase
randomized version of the template, and subtract this
'noise' map from the final score map. For this method
please see STOPGAP as a reference:
https://doi.org/10.1107/S205979832400295X .
--half-precision HALF_PRECISION
Return and save all output in float16 instead of the
default float32
--rng-seed RNG_SEED Specify a seed for the random number generator used
for phase randomization for consistent results!
--relion5-tomograms-star RELION5_TOMOGRAMS_STAR
Here, you can provide a path to a RELION5
tomograms.star file (for example from a tomogram
reconstruction job). pytom-match-pick will fetch all
the tilt-series metadata from this file and overwrite
all other metadata options.
Device control:
-g GPU_IDS [GPU_IDS ...], --gpu-ids GPU_IDS [GPU_IDS ...]
GPU indices to run the program on.
Logging/debugging:
--log LOG Can be set to `info` or `debug`
3. Extracting particles
Both scripts run on the job file created in pytom_match_template.py which contains details about correlation statistics and the output files. The job file always has the format [TOMO_ID]_job.json.
IMPORTANT For both scripts the [-r, --radius-px] option needs to be considered carefully. The particle extraction will mask out spheres with this radius around each peak in the score volume and prevents selecting the same macromolecule twice. It is specified as an integer number of pixels (not Angstrom!) and ideally it should be the radius of the particle of interest. It can be found by dividing the particle radius by the pixel size, e.g. a ribosome (r = 290Å / 2) in a 15Å tomogram should gets a pixel radius of 9.6. As it needs to be an integer value and ribosomes are not perfect spheres, it is best to round it down to 9 pixels.
pytom_extract_candidates.py
STAR file metadata
Resulting STAR files from extraction have three colums with extraction statistics (LCCmax, CutOff, SearchStd). Dividing the LCCmax and the CutOff by the SearchStd, will express them as a number of \(\sigma\) or (3D SNR; similar to Rickgauer et al. (2017).
STAR files written out by the template matching module will have RELION compliant column headers, i.e. rlnCoordinateX and rlnAgleRot, to simplify integration with other software. The Euler angles that are written out therefore also follow the same conventions as RELION and Warp, i.e. rlnAngleRot, rlnAngleTilt, rlnAnglePsi are intrinsic clockwise ZYZ Euler angles. Hence they can be directly used for subtomogram averaging in RELION. See here for more info: https://www.ccpem.ac.uk/user_help/rotation_conventions.php.
Please see the For developers section for more details on the metadata.
Default true positive estimation
The particle extraction has been updated to use the formula in Rickgauer et al. (2017) for finding the extraction threshold based on the false alarm rate. This was not yet described in our IJMS publication but is essentially very similar to the Gaussian fit that we used. However, it is more reliable and also specific to the standard deviation \(\sigma\) of the search in each tomogram. pytom_match_template.py keeps track of \(\sigma\) and stores it in the job file. The user can specify a number of false positives to allow per tomogram with a minimum value of 1. It can be increased to make the extraction threshold more lenient which might increase the number of true positives at the expense of more false positives. The parameter should roughly correspond to the number of false positives that end up in the extracted particle list.
Template matching has a huge search space \(N_{voxels} * N_{rotations}\) which is mainly false positives, and has in comparison a tiny fraction of true positives. If we have a Gaussian for the background (with expected mean 0 and some standard deviation), the false alarm rate can be calculated for a certain cut-off value, as it is dependent on the size of the search space. For example, a false alarm rate of \((N_ {voxels} * N_{rotations})^{-1}\), indicates it would expect 1 false positive in the whole search. This can be calculated with the error function,
, where theta is the cut-off, sigma the standard deviation of the Gaussian, and N the search space. \(n_{\text{FP}}\) represents the scaling by the user of tolerated number of false positives.
Tophat transform filter
This option can be used to filter the score map for sharp peaks (steep local maxima) which usually correspond to true positives. This will be described in a forthcoming publication. For now, you can check out Marten's poster at CCPEM that shows some preliminary results: 10.5281/zenodo.13165643.
usage: pytom_extract_candidates.py [-h] -j JOB_FILE
[--tomogram-mask TOMOGRAM_MASK]
[--ignore_tomogram_mask IGNORE_TOMOGRAM_MASK]
-n NUMBER_OF_PARTICLES
[--number-of-false-positives NUMBER_OF_FALSE_POSITIVES]
-r RADIUS_PX [-c CUT_OFF]
[--tophat-filter TOPHAT_FILTER]
[--tophat-connectivity TOPHAT_CONNECTIVITY]
[--relion5-compat RELION5_COMPAT]
[--log LOG] [--tophat-bins TOPHAT_BINS]
[--plot-bins PLOT_BINS]
Run candidate extraction. -- Marten Chaillet (@McHaillet)
options:
-h, --help show this help message and exit
-j JOB_FILE, --job-file JOB_FILE
JSON file that contain all data on the template
matching job, written out by pytom_match_template.py
in the destination path.
--tomogram-mask TOMOGRAM_MASK
Here you can provide a mask for the extraction with
dimensions (in pixels) equal to the tomogram. All
values in the mask that are smaller or equal to 0 will
be removed, all values larger than 0 are considered
regions of interest. It can be used to extract
annotations only within a specific cellular region. If
the job was run with a tomogram mask, this file will
be used instead of the job mask
--ignore_tomogram_mask IGNORE_TOMOGRAM_MASK
Flag to ignore the input and TM job tomogram mask.
Useful if the scores mrc looks reasonable, but this
finds 0 particles to extract
-n NUMBER_OF_PARTICLES, --number-of-particles NUMBER_OF_PARTICLES
Maximum number of particles to extract from tomogram.
--number-of-false-positives NUMBER_OF_FALSE_POSITIVES
Number of false positives to determine the false alarm
rate. Here one can increase the recall of the particle
of interest at the expense of more false positives.
The default value of 1 is recommended for particles
that can be distinguished well from the background
(high specificity). The value can also be set between
0 and 1 to make the cut-off more restrictive.
-r RADIUS_PX, --radius-px RADIUS_PX
Particle radius in pixels in the tomogram. It is used
during extraction to remove areas around peaks
preventing double extraction.
-c CUT_OFF, --cut-off CUT_OFF
Override automated extraction cutoff estimation and
instead extract the number-of-particles down to this
LCCmax value. Setting to 0 will keep extracting until
number-of-particles, or until there are no positive
values left in the score map. Values larger than 1
make no sense as the correlation cannot be higher than
1.
--tophat-filter TOPHAT_FILTER
Attempt to filter only sharp correlation peaks with a
tophat transform
--tophat-connectivity TOPHAT_CONNECTIVITY
Set kernel connectivity for ndimage binary structure
used for the tophat transform. Integer value in range
1-3. 1 is the most restrictive, 3 the least
restrictive. Generally recommended to leave at 1.
--relion5-compat RELION5_COMPAT
Write out centered coordinates in Angstrom for
RELION5.
--log LOG Can be set to `info` or `debug`
--tophat-bins TOPHAT_BINS
Number of bins to use in the histogram of occurences
in the tophat transform code (for both the estimation
and the plotting).
--plot-bins PLOT_BINS
Number of bins to use for the occurences vs LCC_max
plot.
pytom_estimate_roc.py
This script runs the Gaussian fit as described in the IJMS publication. It requires installation with plotting dependencies as it writes out or displays a figure showing the Gaussian fit and estimated ROC curve. The benefit is that it estimates some classification statistics (such as false discovery rate and sensitivity). You can use it to esimate an extraction threshold for a representative tomogram and then supply this threshold as the [-c, --cut-off] parameter for pytom_extract_candidates.py.
usage: pytom_estimate_roc.py [-h] -j JOB_FILE -n NUMBER_OF_PARTICLES -r
RADIUS_PX [--bins BINS]
[--gaussian-peak GAUSSIAN_PEAK]
[--force-peak FORCE_PEAK] [--crop-plot CROP_PLOT]
[--show-plot SHOW_PLOT] [--log LOG]
[--ignore_tomogram_mask IGNORE_TOMOGRAM_MASK]
Estimate ROC curve from TMJob file. -- Marten Chaillet (@McHaillet)
options:
-h, --help show this help message and exit
-j JOB_FILE, --job-file JOB_FILE
JSON file that contain all data on the template
matching job, written out by pytom_match_template.py
in the destination path.
-n NUMBER_OF_PARTICLES, --number-of-particles NUMBER_OF_PARTICLES
The number of particles to extract and estimate the
ROC on, recommended is to multiply the expected number
of particles by 3.
-r RADIUS_PX, --radius-px RADIUS_PX
Particle radius in pixels in the tomogram. It is used
during extraction to remove areas around peaks
preventing double extraction.
--bins BINS Number of bins for the histogram to fit Gaussians on.
--gaussian-peak GAUSSIAN_PEAK
Expected index of the histogram peak of the Gaussian
fitted to the particle population.
--force-peak FORCE_PEAK
Force the particle peak to the provided peak index.
--crop-plot CROP_PLOT
Flag to crop the plot relative to the height of the
particle population.
--show-plot SHOW_PLOT
Flag to use a pop-up window for the plot instead of
writing it to the location of the job file.
--log LOG Can be set to `info` or `debug`
--ignore_tomogram_mask IGNORE_TOMOGRAM_MASK
Flag to ignore the TM job tomogram mask. Useful if the
scores mrc looks reasonable, but this finds 0
particles
4. Merging annotations for export to other software
After running template matching and candidate extraction on multiple tomograms, each tomogram will have an individual starfile with particle annotations. Each starfile will contain the MicrographName column which refers back to the tomogram name. Multiple starfiles can therefore be appended to results in a large list which can be used in other software (such as RELION, WarpM) to load annotations. These software will link the annotations to specific tilt-series using the MicrographName column.
pytom_merge_stars.py
Without providing any parameters the script will try to merge all the starfiles in the current working directory and save them to a new file particles.star.
usage: pytom_merge_stars.py [-h] [-i INPUT_DIR] [-o OUTPUT_FILE] [--log LOG]
Merge multiple star files in the same directory. -- Marten Chaillet
(@McHaillet)
options:
-h, --help show this help message and exit
-i INPUT_DIR, --input-dir INPUT_DIR
Directory with star files, script will try to merge
all files that end in '.star'.
-o OUTPUT_FILE, --output-file OUTPUT_FILE
Output star file name.
--log LOG Can be set to `info` or `debug`