BactoML - Scikit-learn compatible Machine Learning libs for Online Flowcytometry

• Marie Sadler, Octave Martin, Eliane Roosli, Ilaria Ricchi, Douglas Watson
• bNovate SA, 2018-2020

Overview

This python package contains all of our custom machine learning tools for online flow cytometry. They are built for maximum interoperability with Scikit-learn, so the APIs should be familiar to most. We provide several modules related to fingerprinting:

• Data loading: loading of FCS files
• Data generation: generation of FCS files with artificial contaminations
• Pre-processing: tlog transforms, gating to remove noise
• Fingerprinting: gating to count cells, probability binning
• Estimators: Chi-square risk score estimator.
• Full pipeline: Microbial Community Change Detection (MCCD) algorithm, see tutorial/MCCD_pipeline_example.py

Getting started

Create a virtual environment. We recommend using conda to ensure you have the correct python version:

conda create --name bactoml python=3.6 --no-default-packages
conda activate bactoml

Note: You must run the last line (conda activate bactoml) every time you open a new shell to work on this package!

Install the package with:

pip install -r requirements.txt
python setup.py install

[We use pip inside of conda, because we can't get all packages through conda, so it's easier to use only pip]

Then work through the examples in the examples/ folder.

For developers: again I recommend working in a virtual environment or Anaconda environment. Install the library in development mode in your virtualenv:

pip install -r requirements-test.txt
python setup.py install
python setup.py develop

Then create a branch and hack away.

Testing

Tests use pytest. Install py.test:

pip install pytest

And run:

pytest

Further documentation

Documentation is in the docs/ folder, and is managed by sphinx. To build the documentation, make sure you have installed the developer requirements, then:

cd docs/
make html

The output is in docs/html/_build/index.html

Document your code in numpydoc format. When working on the sphinx rst files, you may find it useful to automatically build the docs every time you save a file:

./scripts/auto_build_doc

Contribution guidelines

See CONTRIBUTING.md.