Inference and interactive interface

The inference stage applies selected MPMA-B or MPMA-E workflows to new microbiome profiles. Run evaluation first; MPMA-E inference also requires a completed ensemble stage.

In a sweep configuration, set:

INFERENCE = mll.Inference(
    abundance_path="new_profiles.tsv",
    targets=("mpma_b", "mpma_e"),
    feature_policy="strict",
)

Then run:

mllabiome path/to/config.py --stage inference

feature_policy="strict" requires matching features; "zero_fill" is an optional alternative. Fitted preprocessing and learner pipelines are reused for inference.

The repository also includes an optional local interactive inference interface  for exploring predictions from completed runs. It is separate from the Python package.

Research use only. Not a medical device. Do not enter personal or patient data into a demo interface.