Current mllabiome package

The current package is the recommended implementation for new analyses when the priority is lightweight configuration, faster iteration, reproducible run artefacts, and out-of-fold explainability in the standard workflow.

For reference, earlier documentation and the legacy implementation are available separately: legacy docs and CMG-GUTS/mllabiome-l .

git clone https://github.com/CMG-GUTS/mllabiome.git
cd mllabiome
uv venv
source .venv/bin/activate
uv pip install -e .

from mllabiome import mll

Workflow

configured MPMA sweep → ensemble sweep → out-of-fold explainability → report

A sweep is defined as an ordinary Python configuration file. The configuration declares the data, taxonomic-resolution sets, abundance transformations, learners, evaluation protocol, ensemble search, and explainability targets. Example configurations in the repository are paired with lightweight data under examples/data/, so a cloned repository can run the examples directly after editable installation.

mllabiome examples/configs_sweep.py

Stage-specific reruns are supported:

mllabiome path/to/config.py --stage evaluate
mllabiome path/to/config.py --stage ensemble
mllabiome path/to/config.py --stage explain
mllabiome path/to/config.py --stage report

Next steps

• Start with the quickstart for a runnable example.
• Use sweep configuration to define datasets, representations, learners, and evaluation.
• Use learners and transformations to add scikit-learn models, custom estimators, and custom abundance transformations.
• Use workflow and outputs to understand generated files.
• Use OOF explainability for the current explanation mechanism.