Skip to main content
skill CX Codex

tooluniverse-microbiome-research — Claude Code skill: research

tooluniverse-microbiome-research is a Claude Code skill for finding microbiome studies, sample taxonomy, strain data, and disease links. It supports literature reviews and culture questions.

Apache-2.0 155 installs ★ 1,798 updated 2026-10-24

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓
npx skills add mims-harvard/ToolUniverse --skill tooluniverse-microbiome-research -a claude-code
Open on GitHub

What tooluniverse-microbiome-research does

This skill connects research questions to MGnify, GTDB, ENA, OLS, EuropePMC, BacDive, MediaDive, and GMrepo. It can find studies by biome, retrieve taxonomic profiles and annotations, search genomes and quality metrics, and look up strain phenotypes or culture-medium recipes. GMrepo provides sample counts and health-condition data for gut microbiome species.

Use MGnify biome hierarchies and ENVO terms to explore environments, or search EuropePMC and PubMed for publications. This is for discovery and reference lookups, not a metagenomics analysis workflow. Broad MGnify searches can be slow, and multiword queries may time out; the documentation recommends trying a concise single-keyword query.

What it helps with

  • find studies for a particular biome
  • retrieve an analysis taxonomic profile
  • cross-check a genome against GTDB taxonomy
  • look up a strain’s culture requirements
  • check sample counts for a species and health condition

Example prompts for the agent

  • Find gut microbiome studies and summarize the metadata for a matching study.
  • Retrieve the taxonomic profile and functional annotations for analysis MGYA00793746.
  • Find strains of this species and look up their culture conditions.

How to install tooluniverse-microbiome-research

1Skills CLI (Vercel)

npx skills add mims-harvard/ToolUniverse --skill tooluniverse-microbiome-research -a claude-code

2Manually: put the files in place

~/.claude/skills/tooluniverse-microbiome-research/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/tooluniverse-microbiome-research/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about tooluniverse-microbiome-research

How do I install tooluniverse-microbiome-research in Claude Code?

Run npx skills add mims-harvard/ToolUniverse --skill tooluniverse-microbiome-research -a claude-code. The skill file is at ~/.claude/skills/tooluniverse-microbiome-research/SKILL.md.

What does tooluniverse-microbiome-research do?

It supports microbiome study discovery, taxonomy and genome lookups, and searches for strain phenotypes, culture media, and disease associations.

Is this a metagenomics analysis workflow?

No. This skill focuses on finding studies and reference data; tooluniverse-metagenomics-analysis is the separate analytical workflow.

Source text

SKILL.md · Apache-2.0
---
name: tooluniverse-microbiome-research
description: Microbiome research using MGnify, GTDB, ENA, OLS (ENVO biomes), EuropePMC, BacDive (strain phenotypes), MediaDive (growth media recipes), and GMrepo (gut microbiome disease-association counts). Covers study discovery, taxonomic profiling, host-microbe interaction analysis, biome-by-condition queries, strain culturing requirements, and phenotype/disease prevalence lookups. Use for microbiome study selection, organism-environment associations, clinical-microbiome literature review, and "how do I grow this organism" / "how prevalent is this species in condition X" questions. Distinct from analytical workflow (use tooluniverse-metagenomics-analysis for that).
disable-model-invocation: true
---

# Microbiome Research with ToolUniverse

Comprehensive microbiome analysis using MGnify (EBI metagenomics), GTDB (genome taxonomy), ENA (sequencing data), OLS (ontology lookup for ENVO biomes), and EuropePMC (literature).

## Core Tools

| Tool | Purpose | Auth |
|------|---------|------|
| **MGnify_search_studies** | Find metagenomics studies by biome/keyword | None |
| **MGnify_get_study_detail** | Study metadata, abstract, sample counts | None |
| **MGnify_list_analyses** | List taxonomic/functional analysis outputs for a study | None |
| **MGnify_get_taxonomy** | Taxonomic composition from an analysis | None |
| **MGnify_get_go_terms** | GO functional annotations from an analysis | None |
| **MGnify_get_interpro** | InterPro protein domain annotations | None |
| **MGnify_list_biomes** | Browse MGnify biome hierarchy | None |
| **MGnify_search_genomes** | Search metagenome-assembled genomes (MAGs) | None |
| **MGnify_get_genome** | Genome quality metrics (completeness, contamination) | None |
| **GTDB_search_genomes** | Search bacterial/archaeal genomes by taxonomy | None |
| **GTDB_get_species** | Species cluster details from GTDB | None |
| **GTDB_get_taxon_info** | Taxonomic rank info in GTDB hierarchy | None |
| **GTDB_search_taxon** | Search taxa by partial name across all ranks | None |
| **ENAPortal_search_studies** | Find sequencing studies in ENA. Query format: `description="keyword"` | None |
| **ENAPortal_search_samples** | Find samples with environmental metadata | None |
| **ols_search_terms** | Search ENVO ontology for biome/environment terms | None |
| **EuropePMC_search_articles** | Find microbiome publications | None |
| **PubMed_search_articles** | Literature search (different coverage than EuropePMC) | None |
| **BacDive_search_by_taxon** | List BacDive strain IDs for a genus (+ optional species) | None |
| **BacDive_get_strain** | Curated strain phenotype: morphology, Gram stain, culture temp/media, oxygen tolerance, isolation source | None |
| **MediaDive_search_media** | Find a cultivation medium by name substring | None |
| **MediaDive_get_medium** | Full recipe (ingredients + amounts + prep steps) for a medium | None |
| **MediaDive_search_ingredients** / **MediaDive_get_ingredient** | Look up a medium ingredient's chemical identifiers (CAS/ChEBI/PubChem) | None |
| **GMrepo_search_species** | Gut-microbiome sample counts and phenotype breadth for a species (96,000+ curated samples) | None |
| **GMrepo_get_phenotypes** | List/filter the 350+ health conditions GMrepo has gut-microbiome data for | None |

**For drug-microbiome studies**, also use:
- `PubChem_get_CID_by_compound_name` / `PubChem_get_compound_properties_by_CID` — drug identity
- `CTD_get_chemical_gene_interactions` — drug-gene interactions (e.g., metformin affects 1,175+ genes)
- `kegg_search_pathway` / `kegg_get_pathway_info` — microbial metabolic pathways (butanoate, propanoate)
- `ReactomeAnalysis_pathway_enrichment` — host pathway enrichment for drug-affected genes
- `drugbank_vocab_search` — drug mechanism and targets

> **MGnify tip**: Use concise single-keyword searches (e.g., "metformin") — multi-word queries may timeout. The MGnify API can be slow for broad searches.

## Quick Start

```python
from tooluniverse import ToolUniverse

tu = ToolUniverse()
tu.load_tools()

# 1. Search for gut microbiome studies
studies = tu.run_one_function({
    'name': 'MGnify_search_studies',
    'arguments': {'search': 'gut microbiome', 'size': 5}
})

# 2. Get study details
detail = tu.run_one_function({
    'name': 'MGnify_get_study_detail',
    'arguments': {'study_accession': 'MGYS00006860'}
})

# 3. List analyses for a study
analyses = tu.run_one_function({
    'name': 'MGnify_list_analyses',
    'arguments': {'study_accession': 'MGYS00006860', 'size': 5}
})

# 4. Get taxonomic profile from an analysis
taxonomy = tu.run_one_function({
    'name': 'MGnify_get_taxonomy',
    'arguments': {'analysis_id': 'MGYA00793746'}
})

# 5. Get functional annotations
go_terms = tu.run_one_function({
    'name': 'MGnify_get_go_terms',
    'arguments': {'analysis_id': 'MGYA00793746'}
})
```

## Common Workflows

### Workflow 1: Study Discovery by Environment

Find studies for a specific biome using MGnify's biome hierarchy:

```python
# Browse biome hierarchy
biomes = tu.run_one_function({
    'name': 'MGnify_list_biomes',
    'arguments': {'lineage': 'root:Host-associated:Human', 'depth': 3}
})

# Search studies in a specific biome
studies = tu.run_one_function({
    'name': 'MGnify_search_studies',
    'arguments': {'biome': 'root:Host-associated:Human:Digestive system', 'size': 10}
})

# Look up ENVO ontology terms for environment metadata
envo = tu.run_one_function({
    'name': 'ols_search_terms',
    'arguments': {'query': 'human gut', 'ontology': 'envo', 'rows': 5}
})
```

### Workflow 2: Taxonomic Profiling

Get the microbial composition of a metagenomics sample:

```python
# Get analyses for a study
analyses = tu.run_one_function({
    'name': 'MGnify_list_analyses',
    'arguments': {'study_accession': 'MGYS00006860', 'size': 3}
})

# Get taxonomy for a specific analysis
taxonomy = tu.run_one_function({
    'name': 'MGnify_get_taxonomy',
    'arguments': {'analysis_id': 'MGYA00793746'}
})
# Returns organisms with lineage, abundance counts, and taxonomy rank
```

### Workflow 3: Genome Quality Assessment

Evaluate metagenome-assembled genomes (MAGs):

```python
# Search for genomes from a specific taxon
genomes = tu.run_one_function({
    'name': 'MGnify_search_genomes',
    'arguments': {'taxonomy': 'Faecalibacterium prausnitzii', 'page_size': 5}
})

# Get quality metrics for a genome
genome = tu.run_one_function({
    'name': 'MGnify_get_genome',
    'arguments': {'genome_id': 'MGYG000000001'}
})
# Returns completeness, contamination, N50, genome length, taxonomy

# Cross-reference with GTDB taxonomy
gtdb = tu.run_one_function({
    'name': 'GTDB_search_genomes',
    'arguments': {'operation': 'search_genomes', 'query': 'Faecalibacterium', 'items_per_page': 5}
})
```

### Workflow 4: Functional Annotation

Discover functional potential of a metagenome:

```python
# GO terms from an analysis
go_terms = tu.run_one_function({
    'name': 'MGnify_get_go_terms',
    'arguments': {'analysis_id': 'MGYA00793746'}
})

# InterPro domains
interpro = tu.run_one_function({
    'name': 'MGnify_get_interpro',
    'arguments': {'analysis_id': 'MGYA00793746'}
})
```

### Workflow 5: Literature Integration

Combine metagenomics data with published research:

```python
# Find relevant publications
papers = tu.run_one_function({
    'name': 'EuropePMC_search_articles',
    'arguments': {'query': 'gut microbiome AND Faecalibacterium AND (IBD OR "Crohn")', 'limit': 10}
})

# Find sequencing data in ENA
ena_studies = tu.run_one_function({
    'name': 'ENAPortal_search_studies',
    'arguments': {'query': 'description="gut microbiome 16S"', 'limit': 5}
})
```

### Workflow 6: Strain Phenotype, Culturing Requirements, and Disease-Association Prevalence

Answer "what is this organism actually like and how prevalent is it in condition X" — a question MGnify/GTDB (composition and taxonomy only) can't answer:

```python
# 1. Find BacDive strain IDs for a taxon (genus required, species optional)
strains = tu.run_one_function({
    'name': 'BacDive_search_by_taxon',
    'arguments': {'genus': 'Faecalibacterium', 'species': 'prausnitzii', 'limit': 3}
})
# -> [{'bacdive_id': 159475}, {'bacdive_id': 159476}, {'bacdive_id': 159477}]

# 2. Get the curated phenotype for one strain
strain = tu.run_one_function({
    'name': 'BacDive_get_strain',
    'arguments': {'bacdive_id': 159475}
})
# -> anaerobe, mesophilic (37C), isolated from human feces, grows on
#    "YCFA-MEDIUM (MODIFIED) (DSMZ Medium 1611)" and chopped-meat medium

# 3. Cross-reference the named medium in MediaDive for the actual recipe
medium = tu.run_one_function({
    'name': 'MediaDive_search_media',
    'arguments': {'query': 'YCFA'}
})
# -> medium_id 1611 matches the BacDive culture_media entry exactly
recipe = tu.run_one_function({
    'name': 'MediaDive_get_medium',
    'arguments': {'medium_id': 1611}
})
# -> full solution-by-solution ingredient list with amounts/units and prep steps

# 4. Check GMrepo for how prevalent/associated the species is in human gut studies
prevalence = tu.run_one_function({
    'name': 'GMrepo_search_species',
    'arguments': {'query': 'Faecalibacterium prausnitzii'}
})
# -> present in 21,286 of ~96,000 samples (22.0%), spanning 88 phenotypes

conditions = tu.run_one_function({
    'name': 'GMrepo_get_phenotypes',
    'arguments': {'query': 'Crohn'}
})
# -> Crohn Disease (MeSH D003424): 5,636 samples across 1,440 species, 602 genera
```

**Reading BacDive's `culture_media` field**: it names the medium (e.g. `"YCFA-MEDIUM (MODIFIED) (DSMZ Medium 1611)"`) but gives no recipe — always follow up in MediaDive by searching the medium name to find its `medium_id`, then `MediaDive_get_medium` for the actual formulation. `oxygen_tolerance` can legitimately be an empty array (not curated for that strain) — don't infer aerobe/anaerobe status from an empty list.

**GMrepo scope note**: `presented_samples`/`all_samples` are curated sample COUNTS from public studies, not a claim about true population prevalence — a species absent from GMrepo's phenotype list for a condition may simply not have been studied there yet, not proven absent from that condition's gut microbiome.

## MGnify Biome Hierarchy

Key biome lineages (use `MGnify_list_biomes` to discover others):
- Human gut: `root:Host-associated:Human:Digestive system`
- Human oral/skin: `root:Host-associated:Human:Oral` / `root:Host-associated:Human:Skin`
- Soil: `root:Environmental:Terrestrial:Soil`
- Ocean/Freshwater: `root:Environmental:Aquatic:Marine` / `root:Environmental:Aquatic:Freshwater`
- Wastewater: `root:Engineered:Wastewater`

## Key Identifiers

MGnify: studies=`MGYS*`, analyses=`MGYA*`, genomes=`MGYG*`. ENA studies=`PRJEB*`. GTDB genomes=`GCA_*`. ENVO terms=`ENVO:*` (e.g. ENVO:00002041). BacDive strain IDs and MediaDive medium/ingredient IDs are plain integers (occasionally suffixed, e.g. `"1a"`) with no fixed prefix — always resolve via search first. GMrepo phenotypes use MeSH IDs (e.g. `D003424` for Crohn Disease).

## Reasoning Framework

### Starting Point: Define the Question First

Microbiome analysis starts with: what is the question? LOOK UP DON'T GUESS — always check the study type and sequencing method before interpreting results.

**Decision tree for data type:**
- Community composition (who is there?) → 16S/ITS amplicon → alpha/beta diversity, differential abundance
- Functional potential (what can they do?) → Shotgun metagenomics → MGnify GO terms, InterPro, KEGG pathways
- Active function (what are they doing now?) → Metatranscriptomics → specialized pipelines (not MGnify/GTDB alone)

Before calling any tool, determine which data type the user has via `MGnify_get_study_detail` — the pipeline type (amplicon vs shotgun) determines which analyses are valid. Do not apply 16S diversity metrics to metagenomic data or vice versa.

### Dysbiosis Assessment Strategy

Dysbiosis (microbial imbalance) is context-dependent — there is no universal "healthy" microbiome. LOOK UP DON'T GUESS — compare to study-matched controls, not general population references.

1. **Check alpha diversity**: Reduced Shannon index relative to controls suggests dysbiosis. Use `MGnify_get_taxonomy` to get community profiles, then assess richness and evenness.
2. **Identify keystone taxa shifts**: Loss of known beneficial taxa (e.g., Faecalibacterium, Roseburia in gut) or bloom of pathobionts (e.g., Enterobacteriaceae). LOOK UP taxa roles with `GTDB_get_species` and literature via `EuropePMC_search_articles`.
3. **Functional consequences**: Does taxonomic shift correlate with loss/gain of metabolic pathways? Check `MGnify_get_go_terms` and `MGnify_get_interpro` for the affected samples.
4. **Confounders**: Antibiotics, diet, age, and geography all affect microbiome composition. A dysbiosis claim requires controlling for these factors or acknowledging them as limitations.

### Taxonomic vs Functional Analysis: When to Use Each

- **Taxonomic analysis alone** is sufficient when the question is "which organisms are present?" or "does community composition differ between groups?" Use `MGnify_get_taxonomy` + `GTDB_search_genomes`.
- **Functional analysis** is needed when the question is "what metabolic capabilities differ?" or "why does a taxonomic shift matter?" Use `MGnify_get_go_terms` + `MGnify_get_interpro` + `kegg_search_pathway`.
- **Both together** when linking organisms to functions (e.g., "which taxa drive butyrate production in healthy vs IBD gut?"). Cross-reference taxonomic profiles with functional annotations from the same MGnify analysis.

### Evidence Grading

| Tier | Description | Example |
|------|-------------|---------|
| **T1** | Replicated finding across multiple cohorts with consistent effect | Reduced Faecalibacterium in IBD (>10 independent studies) |
| **T2** | Single well-powered study (n > 100) with appropriate controls | Metformin-associated Akkermansia enrichment in a controlled trial |
| **T3** | Pilot study or observational association, small sample size | Taxonomic shift in n=15 case-control, no validation cohort |
| **T4** | Computational prediction or single-sample observation | Novel MAG with predicted function, no culture confirmation |

### Interpretation Guidance

**Alpha diversity (within-sample)**: Shannon index measures richness and evenness. Higher Shannon (>3.0 for gut) suggests a stable community. Reduced alpha diversity is associated with dysbiosis (IBD, antibiotics). Always compare to study-matched controls — diversity varies by body site, sequencing depth, and geography.

**Beta diversity (between-sample)**: Bray-Curtis (abundance-based) or UniFrac (phylogenetic). PERMANOVA p < 0.05 with R-squared > 0.05 indicates condition-driven clustering. Low R-squared (<0.02) even with significant p suggests the effect is small relative to inter-individual variation. Choose weighted UniFrac when abundant taxa matter most; unweighted when rare taxa are important.

**Taxonomic composition**: Relative abundance at phylum level (Firmicutes/Bacteroidetes ratio) is a coarse indicator; genus- or species-level resolution is preferred. A taxon present at >1% relative abundance in multiple samples is reliably detected. Taxa at <0.1% may be noise or sequencing artifacts. GTDB taxonomy may reclassify NCBI names (e.g., Firmicutes split into multiple phyla).

**Functional profiling**: GO terms and InterPro domains from MGnify reflect the metabolic potential (not necessarily activity) of the community. Enrichment of specific pathways (e.g., butyrate production, LPS biosynthesis) should be interpreted alongside taxonomic data to identify which organisms contribute the functions.

### Synthesis Questions

A complete microbiome report should answer:
1. How does alpha diversity compare between conditions, and is the difference significant?
2. Does beta diversity analysis show condition-driven clustering (PERMANOVA)?
3. Which taxa are differentially abundant, and are they known commensals or pathobionts?
4. What functional pathways are enriched, and which taxa likely drive them?
5. How do findings compare to published studies for the same biome/condition (literature context)?

## Tips

- MGnify study accessions start with `MGYS`, analyses with `MGYA`, genomes with `MGYG`
- Use `MGnify_list_biomes` first to find the correct biome lineage string
- `MGnify_get_taxonomy` returns phylum-level to species-level composition
- GTDB provides standardized bacterial/archaeal taxonomy (differs from NCBI in some lineages)
- For 16S amplicon studies, taxonomy is the primary output; for shotgun metagenomics, both taxonomy and functional annotations are available
- The `size` parameter in MGnify tools controls results per page (max 100)

© mims-harvard · Apache-2.0 · mims-harvard/ToolUniverse

Similar skills

All skills for Science and research →

From the same set136

The whole set →
tooluniverse-variant-functional-annotation tooluniverse-variant-functional-annotation annotates protein variants with ProtVar, ClinVar, gnomAD, CADD, and ClinGen data. Useful for functional context, population evidence, and variant review. CX Codex +1 229installs tooluniverse-stem-cell-organoid tooluniverse-stem-cell-organoid guides stem cell protocols, pluripotency verification, and organoid fidelity scoring using ToolUniverse biomedical tools. CX Codex +1 226installs tooluniverse-noncoding-rna tooluniverse-noncoding-rna guides non-coding RNA queries across miRBase, LNCipedia, and RNAcentral to inspect targets, transcript structures, and disease links. CX Codex +1 225installs tooluniverse-functional-genomics-screens tooluniverse-functional-genomics-screens is a skill for ranking hits from genetic screens (CRISPR-KO, CRISPRi, shRNA) using essentiality, pathway, and clinical data. CX Codex +1 225installs tooluniverse-protein-interactions The tooluniverse-protein-interactions skill analyzes protein interaction networks through STRING, BioGRID, SASBDB, and Complex Portal. It distinguishes physical binding from functional associations such as co-expression and co-regulation. CX Codex +1 155installs tooluniverse-population-genetics tooluniverse-population-genetics is a Claude Code skill for analyzing genetic diversity, allele frequencies across populations, Hardy-Weinberg equilibrium, and evolutionary constraints. CX Codex +1 150installs tooluniverse-gene-regulatory-networks tooluniverse-gene-regulatory-networks guides gene regulatory network analysis using JASPAR binding motifs, Enrichr ChIP-seq targets, GTEx eQTL data, and STRING protein interactions. CX Codex +1 148installs devtu-benchmark-harness devtu-benchmark-harness — a skill for Claude Code and Codex that runs a five-step feedback loop over ToolUniverse tools: execute a benchmark, map failures to skills, diagnose root causes, route fixes, retest. CX Codex +1 148installs

Project sponsors

Companies that keep this analytics open