README

Delumps is an Rust extension module for computing semantic similarity between individuals within C2S2, a tool for phenotype-driven identification of disease subgroups.

Setup

We deploy Delumps to Python Package Index (PyPi), so installation with PIP should be possible:

python3 -m pip install delumps

Install in production mode

Alternatively, Delumps can be installed from sources:

cd crates/delumps  # Change to the extension directory

python3 -m pip install .

Note, the installation may fail if Rust toolchain is not available on your system.

Install in development mode

For package developers, the package is installed into the active virtual environment by running:

maturin develop -E test
# or
python3 -m pip install -e .[test]

The package is installed in editable mode along with the test extras, that include Pytest and Numpy.

The tests are run from the workspace folder:

pytest

Use with Stratiphy

Delumps provides a delumps.SimilarityMatrixCreatorFactory that can create a delumps.SimilarityMatrixCreator to use with Stratiphy.

import delumps

fpath_hpo = '/path/to/hp.json'
fpath_ic_mica = '/path/to/term-pair-similarity.csv.gz'

factory = delumps.SimilarityMatrixCreatorFactory.from_hpo_and_ic_mica(fpath_hpo, fpath_ic_mica)

The factory needs path to a HPO JSON file and an IC MICA file with precomputed information content (IC) for the most informative common ancestor (MICA) term of an HPO term pair.

Then, similarity matrix creator that uses the Phenomizer semantic similarity measure can be configured by running:

phenomizer_smc = factory.create_phenomizer_smc()

phenomizer_smc can be used as a SimilarityMatrixCreator component of the Stratiphy framework.

Release

We deploy Delumps to PyPi:

cd crates/delumps

maturin publish