Compare and cluster 2D class averages of multiple distinct structures from cryo-EM data based on common lines.
Example mrcs and star file provided in /data, corresponding cryo-EM data at EMPIAR-10268.
See manual.pdf for a brief tutorial.
- Default scoring now uses FT of projections, use
-d Realfor real space projections - GUI now includes option to remove nodes from the graph and input desired # of clusters
upcoming updates - Improved scoring and clustering
- GPU support
- Fixed memory error with multiprocessing
- Added option to downscale class averages
-s(--> faster processing) - Added Wasserstein (Earth mover) distance
- Removed Jupyter Notebook (out of date)
Download the software and create a conda environment -
git clone https://github.com/marcottelab/SLICEM.git
conda env create -f environment.yml
conda activate SLICEM
source deactivate #to return to base env
First generate a score file using slicem.py
usage: SLICEM_DEV.py [-h] -i MRC_INPUT -o OUTPATH
[-m {Euclidean,L1,cosine,EMD,correlate}]
[-s SCALE_FACTOR] [-c NUM_WORKERS] [-d {Fourier,Real}]
[-t {full,valid}] [-a ANGULAR_SAMPLING]
compare similarity of 2D class averages based on common lines
optional arguments:
-h, --help show this help message and exit
-i MRC_INPUT, --input MRC_INPUT
path to mrcs file of 2D class averages
-o OUTPATH, --outpath OUTPATH
path for output files
-m {Euclidean,L1,cosine,EMD,correlate}, --metric {Euclidean,L1,cosine,EMD,correlate}
choose scoring method, default Euclidean
-s SCALE_FACTOR, --scale_factor SCALE_FACTOR
scale factor for downsampling. (e.g. -s 2 converts 200pix box --> 100pix box)
-c NUM_WORKERS, --num_workers NUM_WORKERS
number of CPUs to use, default 1
-d {Fourier,Real}, --domain {Fourier,Real}
Fourier or Real space, default Fourier
-t {full,valid}, --translate {full,valid}
indicate size of score vector, numpy convention, default full
-a ANGULAR_SAMPLING, --angular_sampling ANGULAR_SAMPLING
angle sampling for 1D projections in degrees, default 5
command line example
(SLICEM): python slicem.py -i path/to/input.mrc -o path/to/output/ -c 8
alternative example if running remotely
(SLICEM): nohup python slicem.py -i path/to/input.mrc -o path/to/output/ -m L1 -s 2 -c 10 > log.txt &
Next load the scores and class averages into the GUI for clustering. See manual.pdf for more details
python slicem_gui.py
Separating distinct structures of multiple macromolecular assemblies from cryo-EM projections
https://doi.org/10.1016/j.jsb.2019.107416
report as a GitHub issue or email [email protected]