Contact: ryan.dale@nih.gov
Given the position of variants and protein domains within a cDNA, this tool will run many permutations and calculate empirical p-values for the geometric mean of pairwise distances between variants (to assess spatial clustering) as well as for the number of variants falling within each domain.
Requires Python 3.6+ along with the NumPy and pyyaml packages (which are
indicated in the requirements.txt file. These can be installed via
pip or
conda.
E.g. with conda,
conda create -n spatial-clustering --file requirements.txt
conda activate spatial-clustering
Tests, if you'd like to run them, also require the pytest package to be installed. These are also run automatically via GitHub Actions on every push to this repo.
Edit the config.yml file to reflect your cDNA, variants, and domains of
interest (see that file for details).
Run with the following command:
python domain_and_clustering_analysis.py config.yml --permutations 1000
The default number of permutations is 1e6, but this can be adjusted with the
--permutations argument as in the example above.
The result will be an annotated report printed to stdout, e.g.,
length: 2922
permutations: 1000
actual geometric mean pairwise distance: 1.0416633395161507e-05
distance p-value: 0.3806193806193806
number of variants per domain:
{'signal': 0, 'proregion': 1, 'linker1': 1, 'PHM': 4, 'linker2': 3, 'PAL': 7, 'linker3': 0, 'TMD': 0, 'CD': 0}
p-values per domain:
{'signal': 0.2647352647352647, 'proregion': 0.01098901098901099, 'linker1': 0.004995004995004995, 'PHM': 0.6153846153846154, 'linker2': 0.19080919080919082, 'PAL': 0.12787212787212787, 'linker3': 0.5224775224775224, 'TMD': 0.32367632367632365, 'CD': 0.7792207792207793}