Metabolite Cell Communication (bioSkills)

A Claude Code skill for the metabolic layer of cell–cell communication: scoring producing-enzyme against sensor expression to nominate metabolite-mediated crosstalk, while being blunt about how far that inference reaches.


Type	Claude Skill
Supplier	GPTomics bioSkills (community OSS, MIT)
Availability	GA — part of the bioSkills collection
Pricing	Free / OSS (MIT) — MEBOCOST and the alternative back-ends are installed separately (all open source)
Capabilities	Read/Write — Claude runs the skill’s workflow locally (Python), not as an MCP tool
Verified	works · 2026-08-13
Security	cleared · 2026-08-13 — GPTomics/bioSkills MIT, no external credentials

How to install

bioSkills is not an npm package — skills are plain markdown/code read directly by the agent. Clone the repo, then either run the installer for the whole category or copy the single skill directory.

Claude Code — clone and install via the bundled script:
```
git clone https://github.com/GPTomics/bioSkills
cd bioSkills
./install-claude.sh --categories "single-cell"
```
The installer copies matching skills into ~/.claude/skills/ (default target). Use ./install-claude.sh --list to preview the skills first.
Claude Code / other agents — copy just this one skill:
```
cp -r bioSkills/single-cell/metabolite-communication ~/.claude/skills/
```
(run from inside your clone — the previous step left you in bioSkills/; otherwise replace bioSkills/ with the absolute path of your clone, e.g. /Users/you/repos/bioSkills).
Prerequisites — the single-cell base stack:
```
pip install "scanpy>=1.10" "anndata>=0.10"
```
Prerequisite — MEBOCOST 1.0+ (the primary tool). Unverified — a PyPI distribution was not confirmed this run, so install from the upstream repository:
```
pip install "git+https://github.com/zhengrongbin/MEBOCOST.git"
```
Confirm with pip show mebocost. MEBOCOST also needs its species-matched enzyme/sensor database and config files, which ship with the repository — note the path to them, because create_obj() takes it as an argument.

The alternatives the skill routes to (scFEA, Compass, NeuronChat) are separate installs and are only needed if the decision rules below point at them.

What it does

Data preparation — expects a log-normalized AnnData with gene symbols, not Ensembl IDs, cell-type labels, and an optional condition column; ambient RNA should be removed first (SoupX, DecontX, or CellBender).
Object creation — mebocost.create_obj() with a species-matched config and database path.
Inference — infer_commu() scores sender→receiver metabolite pairs using label-permutation testing.
Filtering — keep results at permutation_test_fdr < 0.05; transporter-based calls are separated out as lower-confidence.
Method routing — MEBOCOST for enzyme-to-sensor crosstalk between cell types; scFEA for relative per-cell metabolic flux (support for metabolite reasoning, not communication itself); Compass for comparing metabolic state between conditions via flux-balance analysis — its output measures reaction favourability, not secretion, so it does not license a communication claim; NeuronChat for neural systems only (glutamate, GABA, dopamine, serotonin, neuropeptides).

Working parameters: cutoff_prop = 0.15 (a gene expressed in fewer than 15% of a group is mostly dropout), n_shuffle = 1000 for a stable permutation FDR, min_cell_number = 10 to stop tiny groups inflating scores, and FDR rather than raw p-value for filtering.

Primary use cases: nominating metabolic crosstalk between cell types, predicting metabolite secretion and sensing, choosing among metabolic-communication methods.

Notes

The skill’s central warning is the reason to prefer it over running MEBOCOST unaided: metabolite-mediated communication is a double inference, and the most speculative layer of cell–cell communication analysis. The chain runs enzyme mRNA → protein → activity → flux → metabolite pool → secretion → sensing, and none of those arrows is measured by scRNA-seq. Validation means metabolomics, mass-spectrometry imaging, isotope tracing, or enzyme/sensor perturbation — not a second expression-based method agreeing with the first.

Three failure modes to watch. Bidirectional transporters annotated as “sensors” may export rather than import the metabolite, which is why transporter-based hits are downgraded. Ambient RNA inflates apparent enzyme expression in non-producing cell types, so decontaminate before inference. And a species mismatch — mouse data scored against the human enzyme/sensor database — returns almost nothing, which reads as a negative result rather than a configuration error.

Upstream skill front-matter name is bio-single-cell-metabolite-communication (tool_type: python, primary_tool: MeboCost); upstream directory single-cell/metabolite-communication. Pairs with CellChat and LIANA for the better-established ligand–receptor layer, COBRApy for constraint-based metabolic modelling, scanpy / AnnData for the data objects, and Metabolomics Workbench or HMDB when looking for the measured metabolite evidence the skill asks for.

Sources

Installed this tool?

Share feedback — install path, OS, errors, workarounds. The form opens with this tool pre-selected and a link back to this page.