| CryoAlign2 | efficient global and local Cryo-EM map retrieval based on parallel-accelerated local spatial structural features |
|---|
We developed an alignment-based retrieval tool to perform both global and local retrieval. Our approach adopts parallel-accelerated CryoAlign for high-precision 3D alignment and transforms density maps into point clouds for efficient retrieval and storage. Additionally, a multi-dimension scoring function is introduced to accurately assess structural similarities between superimposed density maps.
Note that CryoAlign2 depends on and uses several external programs and libraries.If you have a Docker environment with GPU, we strongly recommend you to generate image using the dockerfile.
- Ubuntu 18.04 or later
- CMake 3.20+
- Open3D 0.18.0
- Libtorch 2.2.0 CUDA(Version above 12.2)
- FLANN
- PCL(point cloud libraries)
- EIGEN3
- MLPACK 3.2.2
- CNPY
- TEASER++
- FFTW
We store the public release versions of CryoAlign2 on GitHub, a site that provides code-development with version control and issue tracking through the use of git. To clone the repository, run the following command
git clone https://github.com/JokerL2/CryoAlign2.git
For the convenience of environment configuration, we have provided Docker images that include the necessary environments and external libraries. Users must ensure that the external libraries have been properly installed on their servers.
- Using Docker
docker build -f dockerfile -t [image name] .
docker run -it --name [container name] --gpus all [image name]
- Some external libraries and tools
Download libtorch
cd ~/
wget https://download.pytorch.org/libtorch/nightly/cu121/libtorch-cxx11-abi-shared-with-deps-2.2.0.dev20231213%2Bcu121.zip
unzip libtorch-cxx11-abi-shared-with-deps-2.2.0.dev20231213+cu121.zip
- Install CryoAlign2
cd /CryoAlign2 && mkdir build && cd build
cmake ..
make
After installation, three executable files will be generated.
CryoAlign: Full-process density map similarity comparison, including density map sampling, keypoint extraction, and alignment scoring.
CryoAlign_extract_keypoints: Performs density map sampling and keypoint extraction only.
CryoAlign_alignment: Direct alignment scoring using sampled points and keypoints from density maps.
Usage:
CryoAlign [data dir] [source.map] [source contour level] [target.map]
[target contour level] [source.pdb] [source sup.pdb] [voxel_size] [feature_radius] [alg_type]
Options:
-
–data_dir: Map file path. -
–source_map: Source emdb num. -
–source_contour_level: Author recommend contour level. -
–target_map: Target emdb num. -
–target_contour_level: Author recommend contour level. -
–source_pdb(optional): Source pdb name. -
–source_sup_pdb(optional): Transformed source pdb name (ground truth). -
–voxel_size: Sampling interval (defaults 5.0). -
–feature_radius: Radius for feature construction (defaults 7.0). -
–alg_type:Global_alignmentorMask_alignment.
Example:
For Global_alignment:
CryoAlign --data_dir ../../example_dataset/emd_3695_emd_3696/ --source_map EMD-3695.map --source_contour_level 0.008 --target_map EMD-3696.map --target_contour_level 0.002 --source_pdb 5nsr.pdb --source_sup_pdb 5nsr_sup.pdb --voxel_size 5.0 --feature_radius 7.0 --alg_type global
For Mask_alignment:
CryoAlign --data_dir ../../example_dataset/emd_3695_emd_3696/ --source_map EMD-3695.map --source_contour_level 0.008 --target_map EMD-3696.map --target_contour_level 0.002 --source_pdb 5nsr.pdb --source_sup_pdb 5nsr_sup.pdb --voxel_size 5.0 --feature_radius 7.0 --alg_type mask
Usage:
CryoAlign_extract_keypoints [data dir] [source.map] [source contour level] [target.map] [target contour level] [voxel_size]
Options:
-
–data_dir: Map file path. -
–map_name: Source emdb num. -
–contour_level: Author recommend contour level. -
–voxel_size: Sampling interval. (defaults 5.0)
Example:
CryoAlign_extract_keypoints --data_dir ../../example_dataset/emd_3695_emd_3696/ --map_name EMD-3695.map --contour_level 0.008 --voxel_size 5.0
CryoAlign_alignment [data dir] [source_xyz] [target_xyz] [source_sample]
[target_sample] [source.pdb] [source sup.pdb] [voxel_size] [feature_radius]
[alg_type]
Options:
--data_dir: Map file path.
--source_xyz: Source map keypoints file.
--target_xyz: Target map keypoints file
--source_sample: Source map sample file.
--target_sample: Target map sample file.
--source_pdb(optional): Source pdb name.
--source_sup_pdb(optional): Transformed source pdb name (ground truth).
--voxel_size: Sampling interval. (defaults 5.0)
--feature_radius: Radius for feature construction. (defaults 7.0)
--alg_type: Global_alignment or Mask_alignment.
Examples
For Global Alignment
CryoAlign_alignment --data_dir ../../example_dataset/emd_3695_emd_3696/ \
--source_xyz Points_3695_5.00_Key.xyz --target_xyz Points_3696_5.00_Key.xyz \
--source_sample EMD-3695_5.00.txt --target_sample EMD-3696_5.00.txt \
--source_pdb 5nsr.pdb --source_sup_pdb 5nsr_sup.pdb --voxel_size 5.0 \
--feature_radius 7.0 --alg_type global
For Mask Alignment
CryoAlign_alignment --data_dir ../../example_dataset/emd_3695_emd_3696/ \
--source_xyz Points_3695_5.00_Key.xyz --target_xyz Points_3696_5.00_Key.xyz \
--source_sample EMD-3695_5.00.txt --target_sample EMD-3696_5.00.txt \
--source_pdb 5nsr.pdb --source_sup_pdb 5nsr_sup.pdb --voxel_size 5.0 \
--feature_radius 7.0 --alg_type mask
Use CryoAlign_extract_keypoints to build a retrieval database. For the script, see script/CreateDB.py.
python script/CreateDB.py
You will get a key point cloud and sample point cloud library, similar to the database example
Use CryoAlign_alignment to perform the search. See script/CryoSearch.py for the script.
python script/CryoSearch.py
You will get a density map similarity score file res.txt
8553-8551 0.810762
8587-8551 0.7663
8889-8587 0.456922
8889-8551 0.689744
...
Use script/Transform_map.py to overlay the density map, and save the RT (rotation translation) matrix for each alignment.
python script/Transform_map.py