MIMOSA2¶
Coming soon
Runner not wired yet. R-bridge (mimosa2::run_mimosa2()) required
before this module can execute.
MIMOSA2 (Model-based Integration of Metabolite Observations and Species Abundances) is the closest peer to GeMMA's punch framework in the literature: it computes a per-metabolite "Community-level Metabolic Potential" (CMP) score from taxon abundance × per-species reference-genome reactions, then regresses the measured metabolite on CMP to determine which metabolites are explained by the microbiome and which taxa drive the prediction.
Comparator positioning vs GeMMA¶
| Axis | GeMMA (punch) | MIMOSA2 (CMP) |
|---|---|---|
| Modelling layer | Genome-scale metabolic models (AGORA2 / VMH) | Per-species reference reactions (KEGG / AGORA2) |
| Direction of scoring | Taxon-centric (which taxa punch above weight) | Metabolite-centric (which metabolites are microbiome-governed) |
| Classifier categories | driver / passenger / protective / depleted | microbiome-governed / residual / disagreeing |
| Requires metabolomics | Yes (for punch weighting + concordance) | Yes (for CMP-vs-observed regression) |
Running both on the same cohort gives the paper one taxon-first and one metabolite-first view of the same community↔metabolome coupling.
Required inputs¶
- Microbiome abundance table (this dataset).
- Paired metabolomics table (auto-resolved via the project's sibling metabolomics dataset, same mechanism as GeMMA).
- MIMOSA2 reference reaction DB — pulled from AGORA2 / KEGG; already staged by GeMMA, so can be reused without extra downloads.
Planned outputs¶
- Per-metabolite CMP-vs-observed scatter + per-taxon contributor ranking.
- MIMOSA2's microbiome-governed / residual / disagreeing classifier plotted next to GeMMA's driver / passenger / protective / depleted so users can see how both framings label the same samples.