MicrobiomeKG¶

A microbiome knowledge graph on kglite, with the evidence model as the point: every association edge carries its study design, direction, sample sizes and citing paper, and the measured negatives — “tested and nothing happened” — are kept as their own relationships rather than folded away.

Install the package from PyPI:

python -m pip install microbiomekg

Prepare, fetch and build over one data directory:

import microbiomekg as mkg
data = "./my-data"
mkg.prepare(data)                 # create input directories and print what each file needs
mkg.fetch(data, missing=True)     # fetch missing automatic inputs, then report again
result = mkg.build(data)          # build from what is present; no file saved by default

microbiomekg status --data ./my-data --create
microbiomekg fetch  --data ./my-data --missing
microbiomekg build  --data ./my-data
microbiomekg serve --graph graph/microbiomekg.kgl  # the MCP server, read-only

No data and no built graph ships: 26% of evidence-bearing edges have no redistribution permission, so what ships is the pipeline.

Guides

Design notes

Research