← Previous Home Next →

How to use Icecream

Rapid installation, usage and reconstruction

The first goal of CryoLithe is to be ultra-fast. Thus, we have made the installation and usage as simple as possible. For custom usage, please see the details on the Github page.

Installation

conda create -n CryoLithe python=3.9
conda activate CryoLithe
pip3 install torch torchvision torchaudio
pip install git+https://github.com:swing-research/CryoLithe.git

Download the trained models

cryolithe download

Reconstruction

cryolithe reconstruct \ 
    --proj-file path_projections \
    --angle-file path_angles \
    --save-dir ./CryoLithe_results/ \
    --save-name output_volume.mrc \
    --device 0 \
    --n3 256 \ 
    --batch-size 100000  # Depends on the memory of your GPU

where path_projections and path_angles are the paths to the projection (mrc or st file) and angle files (tlt or txt) of your tilt-series.

How to obtained aligned tilt-series

CryoLithe requires aligned tilt-series to perform reconstruction. Raw tilt-series come unaligned, and it can be challenging to perform such alignment in general. We provide a simple approach relying on AreTomo 3. This method should provide decent results for most common dataset. We assume that AreTomo has been installed, see the dedicated webpage if needed. We also assume that Imod is available, see the dedicated webpage.

Get data to test on. We take the first tilt-series from EMPIAR 11830, but feel free to use your own data.

mkdir data
wget https://ftp.ebi.ac.uk/empiar/world_availability/11830/data/chlamy_visual_proteomics/01082023_BrnoKrios_Arctis_WebUI_Position_1/01082023_BrnoKrios_Arctis_WebUI_Position_1.st -P ./data/
wget https://ftp.ebi.ac.uk/empiar/world_availability/11830/data/chlamy_visual_proteomics/01082023_BrnoKrios_Arctis_WebUI_Position_1/01082023_BrnoKrios_Arctis_WebUI_Position_1.rawtlt -P ./data/

Align using AreTomo 3

mkdir data/AreTomo
AreTomo3 -InPrefix data/01082023_BrnoKrios_Arctis_WebUI_Position_1 -InSuffix st -OutDir ./data/AreTomo/ -Cmd 0 -VolZ 0 -AlignZ 256 -AtBin 4 -Gpu 0 -CorrCTF 1 -PixSize 1.9 -Kv 300 -Cs 2.7 -Serial 1  -DarkTol 0  -TiltCor 1 -FlipVol 0 -OutImod 3

Align back. This requires the cryoet-alignment package that can be installed using pip.

pip install cryoet-alignment
cryolithe AreTomoToImod --aln-path ./data/AreTomo/01082023_BrnoKrios_Arctis_WebUI_Position_1.aln --output-path ./data/01082023_BrnoKrios_Arctis_WebUI_Position_1
newstack -input data/01082023_BrnoKrios_Arctis_WebUI_Position_1.st -output data/01082023_BrnoKrios_Arctis_WebUI_Position_1_aligned.st -xf ./data/01082023_BrnoKrios_Arctis_WebUI_Position_1.xf -bin 4

Finally, CryoLithe’s reconstruction can be obtained:

cryolithe reconstruct --proj-file data/01082023_BrnoKrios_Arctis_WebUI_Position_1_aligned.st --angle-file data/01082023_BrnoKrios_Arctis_WebUI_Position_1.rawtlt --save-dir ./data/ --save-name 01082023_BrnoKrios_Arctis_WebUI_Position_1_CryoLithe.mrc  --device 0 --n3 256 --batch-size 100000  --pixel 0# Depends on the memory of your GPU

Re-train on your own data

Precise procedure cooming soon!

Parameters of the config file

The default config file can be found here. The different possible parameters are reported below with a brief description.

    data
Key Default Description
model_dir null Path to trained model directory.
proj_file null Path to projections .mrc/.mrcs file.
angle_file null Path to tilt-angle file.
save_dir null Directory to write reconstruction.
save_name null Output volume filename.
    train_params
Key Default Description
patch_scale 1 Override model patch scale. The only parameter that will influence the reconstruction quality. Default is 1. Greater than one means that effective field of the patch increases, and lower than one means that it decreases.
device 0 GPU number or device name.
downsample_projections False Whether to downsample the input projections for reconstruction. If true, the projections will be downsampled by the specified factor.
downsample_factor 0.25 Factor by which to downsample the input projections if downsample_projections is true.
anti_alias true Whether to apply an anti-aliasing filter to the projections before downsampling. This can help reduce artifacts in the reconstruction when downsampling.
N3 256 Volume size along z-axis, after downsampling.
batch_size 100000 Batch size for point inference. Reduce if memory issue.
num_workers 0 Number of worker processes for data loading. Set to 0 to use the main process.
model_variant cryolithe Use cryolithe-pixel (longer, slighlty better quality) or cryolithe (faster)..
← Previous Home Next →