Skip to main content

Clusters conformations of monomeric protein

Project description

Monomeric protein conformational state clustering

These scripts can be used to cluster a parsed set of monomeric protein chains via a global conformational change metric based on CA distances. Once the polypeptide chains destined for clustering have been specified, a pairwise CA distance matrix for each chain is produced. Distance difference matrices are then generated, again, pairwise but between CA distance matrices. Therefore, for N unique peptide chains, N CA distance matrices and N*(N-1)/1 distance difference matrices are generated. NB: the score between A->B is the same as B->A.

Additional scripts are provided to cluster the chains based on distance-based scores calculated from all pairwise distance difference matricies, as well as scripts to produce dendrograms of the clustering results, and heatmaps for each distance difference matrix.

Example input data is provided in the benchmark_data/examples folder, including scripts to download and save data from the PDBe-KB's benchmark conformational state dataset. Example scripts are included in examples, which run complete executions of the entire pipeline for a selection of structures from several difference UniProt accessions.

For intructions on importing protein-cluster-conformers into your own Python code, refer to /tutorials/instructions.ipynb.

Dependencies:

protein-cluster-conformers requires >=Python3.10 to run. Initialise virtual environment and install dependencies with:

cd protein-cluster-conformers
python3.10 -m venv cluster_venv
source cluster_venv/bin/activate
pip install -r requirements.txt

CLI: Clustering structures

To cluster a set of protein structures, run the find_clusters.py script:

python3 find_conformers.py [-h] [-v] -u UNIPROT -m MMCIF [MMCIF ...]
							[-s PATH_CLUSTERS] -c PATH_CA [-d PATH_DD]
                          	[-g PATH_DENDROGRAM [PATH_DENDROGRAM ...]]
                          	[-w PATH_SWARM [PATH_SWARM ...]] [-o PATH_HISTOGRAM]
                          	[-a PATH_ALPHA_FOLD]
                            [-0 FIRST_RESIDUE_POSITION] [-1 LAST_RESIDUE_POSITION]

The following parameters can be parsed:

required arguments:
  -u UNIPROT, --uniprot UNIPROT
                        UniProt accession
  -m MMCIF [MMCIF ...], --mmcif MMCIF [MMCIF ...]
                        Enter list of paths to mmCIFs that overlap a given UniProt segment
optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         Increase verbosity
  -s PATH_CLUSTERS, --path_clusters PATH_CLUSTERS
                        Path to save clustering results
  -c PATH_CA, --path_ca PATH_CA
                        Path to save CA distance matrices
  -d PATH_DD, --path_dd PATH_DD
                        Path to save distance difference matrices
  -g PATH_DENDROGRAM [PATH_DENDROGRAM ...], --path_dendrogram PATH_DENDROGRAM [PATH_DENDROGRAM ...]
                        Path to save dendrogram of clustering results
  -w PATH_SWARM [PATH_SWARM ...], --path_swarm PATH_SWARM [PATH_SWARM ...]
                        Path to save swarm plot of scores
  -o PATH_HISTOGRAM, --path_histogram PATH_HISTOGRAM
                        Path to save histograms of distance difference maps
  -a PATH_ALPHA_FOLD, --path_alpha_fold PATH_ALPHA_FOLD
                        Path to save AlphaFold Database structure
  -0 FIRST_RESIDUE_POSITION, --first_residue_position FIRST_RESIDUE_POSITION
                        First residue position in (UniProt) sequence
  -1 LAST_RESIDUE_POSITION, --last_residue_position LAST_RESIDUE_POSITION
                        Last residue position in (UniProt) sequence

Run instructions

Option 1) Cluster and save matrices

To only cluster a set of monomeric protein structures that share part or all of the same UniProt sequence, run:

python3 find_clusters.py -u "A12345" \
    -m /path/to/structure_1.cif [chains] \
    -m /path/to/structure_2.cif [chains] \
    ... \
    -m /path/to/structure_N.cif [chains] \
    -s /path/to/save/clustering/results/

The paths to each structure are parsed using the -m flag.

Chain IDs (only struct_asym_id is currently recognised at the moment) should be given as space-delimited arguments after the path. Parse in multiple structures using consecutive -m flags. The UniProt accession must be parsed using the -u flag.

Example: O34926

python3 find_conformers.py -u "O34926" \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -s benchmark_data/examples/O34926/O34926_cluster_results/

By default, the pipeline only clusters the parsed mmCIFs (and specified chains), saving clustering results to a CSV file in -s specified directory.


Option 2) Save CA matrices only

To save the matrices produced in the pipeline, simply specify the path in which to save them using the -c flag for CA distance matrices and the -d flag for CA distance difference matrices:

$ python find_conformers.py -u "A12345" \
    -m /path/to/structure_1.cif [chains] \
    -m ... \
    -s /path/to/save/cluster_results.csv \
    -c /path/to/save/CA/distance/matices \
    -d /path/to/save/distance/difference/matrices/

Example: O34926

python3 find_conformers.py -u "O34926" \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances \

Option 3) Render distance difference maps only

2D histograms (heatmaps) can be rendered and saved for each CA distance difference matrix by specifying the save directory using the -o flag:

$ python find_conformers.py -u "A12345" \
    -m /path/to/structure_1.cif [chains] \
    -m ... \
    -o /path/to/save/distance/difference/2D/histograms/

The resulting plots are saved in PNG format (to save render time). E.g:

Distance difference map of 6hac chain A to 6hae chain K

Example: O34926

python3 find_conformers.py -u "O34926" \
	-m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -o benchmark_data/examples/O34926/O34926_distance_difference_maps/

Option 4) Render dendrogram only

From the clustering results, a dendrogram can be rendered to show the relationships between all clustered chains. To save a dendrogram of the hierarchical clustering results, run:

$ python find_conformers.py -u "A12345" \
    -m /path/to/structure_1.cif [chains] \
    -m ... \
    -g /path/to/save/dendrogram/ [png svg]

where either a png or svg file type is saved. E.g.

Dendrogram of clustered UniProt:P14902 chains, via UPGMA agglomerative clustering Dendrogram of clustered UniProt:P14902 chains, via UPGMA agglomerative clustering

Example: O34926

python3 find_conformers.py -u "O34926" \
	-m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -g benchmark_data/examples/O34926/O34926_cluster_results/ png svg

Option 6) Include AlphaFold Database structure when generating CA and distance difference matrices

By parsing in the -a flag, the script will attempt to download and cluster the pre-generated AlphaFold structure, stored on the AlphaFold Database. You do not need to have downloaded the predicted AlphaFold structure already but must be connected to the internet. The structure will be saved

$ python find_conformers.py -u "A12345" \
		-m /path/to/structure_1.cif [chains] \
    -m ... \
    -a

Example: O34926

python3 find_conformers.py -u "O34926" \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -a benchmark_data/examples/O34926/O34926_path_alphafold/

Option 7) Run all

Example #1: O34926

python3 find_conformers.py -u "O34926" \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances/ \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -s benchmark_data/examples/O34926/O34926_cluster_results/ \
    -o benchmark_data/examples/O34926/O34926_distance_difference_maps/ \
    -g benchmark_data/examples/O34926/O34926_cluster_results/ png svg \
    -a benchmark_data/examples/O34926/O34926_path_alphafold/

or use the examples/run_O34926.sh script.

$ ./examples/run_O34926.sh

Example #2: P15291

python3 find_conformers.py -u "P15291" \
    -m benchmark_data/examples/P15291/P15291_updated_mmcif/2fy7_updated.cif A \
    -m benchmark_data/examples/P15291/P15291_updated_mmcif/2fya_updated.cif A \
    -m benchmark_data/examples/P15291/P15291_updated_mmcif/2fyb_updated.cif A \
    -m benchmark_data/examples/P15291/P15291_updated_mmcif/6fwu_updated.cif A B \
    -m benchmark_data/examples/P15291/P15291_updated_mmcif/2fyc_updated.cif A B \
    -m benchmark_data/examples/P15291/P15291_updated_mmcif/2fyd_updated.cif A B \
    -c benchmark_data/examples/P15291/P15291_ca_distances/ \
    -d benchmark_data/examples/P15291/P15291_distance_differences/ \
    -s benchmark_data/examples/P15291/P15291_cluster_results/ \
    -o benchmark_data/examples/P15291/P15291_distance_difference_maps/ \
    -g benchmark_data/examples/P15291/P15291_cluster_results/ png svg \
    -a benchmark_data/examples/P15291/P15291_path_alphafold/

or execute the examples/run_P15291.sh script.

$ ./examples/run_P15291.sh

When imported into Orc, the arguments required to execute the clustering process correctly will be parsed into the class instance and related methods as lists generated from the preceding functions called by the existing protein-superpose pipeline.

Optional arguments

The start and end residue positions can be parsed into the script using the -0 and -1 flags, respectively. This will not restrict the residue ranges during clustering but will be used to label the axes of the distance difference maps and dendrograms.

Example: O34926:

python3 find_conformers.py -u "O34926" \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances/ \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -s benchmark_data/examples/O34926/O34926_cluster_results/ \
    -g benchmark_data/examples/O34926/O34926_cluster_results/ png svg \
    -0 1 \
    -1 405

The pipeline will avoid re-processing existing files where it files them. To update a single PDB entry, specify the PDB accession using the -i flag, e.g. -i 3nc3. To force all entries to be re-processed, use the -f flag, which will overwrite existing files indescriminately.

Example: O34926:

python3 find_conformers.py -u "O34926" \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc3_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc5_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc6_updated.cif A B \
    -m benchmark_data/examples/O34926/O34926_updated_mmcif/3nc7_updated.cif A B \
    -c benchmark_data/examples/O34926/O34926_ca_distances/ \
    -d benchmark_data/examples/O34926/O34926_distance_differences/ \
    -s benchmark_data/examples/O34926/O34926_cluster_results/ \
    -g benchmark_data/examples/O34926/O34926_cluster_results/ png svg \
    -f

Run on benchmark dataset

The scripts above are called by the run_benchmark.py wrapper. To generate conformational clustering results for the included benchmark dataset, run:

python3 cluster_benchmark.py

This will call the run_benchmark(...) functions included in ca_distance.py, distance_difference.py, cluster.py, plot_distance_difference.py, plot_dendrogram.py and plot_swarm_plot.py. No arguments need parsing into the script.

Results will be saved in the ./benchmark_data/ folder.


Contributing

Install developer dependencies:

pip install -r dev-requirements.txt

To run unit tests on the package, the Pytest framework is recommented and can be performed with:

pytest --cov=cluster_conformers --cov-report=html -v

The following dependencies will be required:

  • pytest-cov
  • pytest-forked
  • purest-xdist (optional)

They are installed along with the main package dependencies in requirements.txt.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

protein-cluster-conformers-1.2.1.tar.gz (47.4 kB view details)

Uploaded Source

Built Distribution

protein_cluster_conformers-1.2.1-py3-none-any.whl (55.0 kB view details)

Uploaded Python 3

File details

Details for the file protein-cluster-conformers-1.2.1.tar.gz.

File metadata

File hashes

Hashes for protein-cluster-conformers-1.2.1.tar.gz
Algorithm Hash digest
SHA256 e014a89407fbdf03f691bcb9d84115fe870f3c67579faf66a779cf03e9ba0460
MD5 93dc564c482145f817420e0d04990849
BLAKE2b-256 40246c6c535ec1e18e2d9dd016c75c3ce4222419b99d30596ad0156ed8cb65db

See more details on using hashes here.

File details

Details for the file protein_cluster_conformers-1.2.1-py3-none-any.whl.

File metadata

File hashes

Hashes for protein_cluster_conformers-1.2.1-py3-none-any.whl
Algorithm Hash digest
SHA256 a94d185612551c078c0eceab88c2a56af8f0bce34f06da54cae0220373bd4bc4
MD5 985ee3c4448acff02d46cb8c160093e3
BLAKE2b-256 ff70e738a0338f7d725f8ef93932e98702cf6f995b6460a0c0a71ab09d1aac57

See more details on using hashes here.

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page