Skip to main content
skill CX Codex

tooluniverse-pathway-disease-genetics — Claude Code skill: setup

tooluniverse-pathway-disease-genetics traces GWAS variants through causal genes and biological pathways to drug targets. Use it to study disease mechanisms or prioritize repurposing candidates.

Apache-2.0 239 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-pathway-disease-genetics -a claude-code
Open on GitHub

What tooluniverse-pathway-disease-genetics does

This skill builds a disease-genetics workflow from GWAS signals to candidate genes, pathways, and drugs. It can resolve a disease with OpenTargets, collect variants with GWAS tools, annotate them through Ensembl VEP, and inspect tissue-specific eQTL evidence with GTEx.

Pathway analysis can compare Reactome, KEGG, STRING, and PANTHER results. DGIdb and OpenTargets add druggability categories, target evidence, and associated approved or clinical drugs. When scoring or processing data, the skill is designed to run Python with pandas, scipy, statsmodels, and matplotlib. gwas_get_associations_for_trait is broken, so the workflow uses gwas_search_associations; MetaCyc is currently unavailable.

The workflow treats non-coding variants with care: the nearest gene may not be the causal one. Tissue-relevant eQTL support and repeated pathway signals across databases are used to rank hypotheses, rather than relying on a single annotation.

What it helps with

  • collect GWAS hits for a disease
  • trace a variant to a causal gene
  • test genes for pathway enrichment
  • check whether targets are druggable
  • find drugs linked to GWAS-supported targets

Example prompts for the agent

  • Map GWAS variants for type 2 diabetes to causal genes and biological pathways.
  • Check GTEx eQTL evidence for each candidate in the disease-relevant tissue.
  • Find approved or clinical drugs linked to genes supported by the GWAS results.

How to install tooluniverse-pathway-disease-genetics

1Skills CLI (Vercel)

npx skills add mims-harvard/ToolUniverse --skill tooluniverse-pathway-disease-genetics -a claude-code

2Manually: put the files in place

~/.claude/skills/tooluniverse-pathway-disease-genetics/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/tooluniverse-pathway-disease-genetics/

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

Questions about tooluniverse-pathway-disease-genetics

How do I install tooluniverse-pathway-disease-genetics in Claude Code?

Run npx skills add mims-harvard/ToolUniverse --skill tooluniverse-pathway-disease-genetics -a claude-code, then check ~/.claude/skills/tooluniverse-pathway-disease-genetics/SKILL.md.

What does tooluniverse-pathway-disease-genetics do?

It connects GWAS variants with candidate causal genes, biological pathways, and drug targets. It also examines eQTL evidence in disease-relevant tissues.

Which pathway and drug databases does it use?

Pathway work can use Reactome, KEGG, STRING, and PANTHER. DGIdb and OpenTargets are used for druggability and target-drug evidence.

Source text

SKILL.md · Apache-2.0
---
name: tooluniverse-pathway-disease-genetics
description: Connect GWAS variants to biological pathways and druggable targets. Maps GWAS hits to causal genes (via fine-mapping/eQTL), then to pathways (Reactome, KEGG, WikiPathways), then to existing drugs hitting those pathways. Use for pathway-level disease mechanisms, druggable-pathway prioritization from GWAS, SNP-to-pathway-to-target tracing, and tissue-specific eQTL evidence for drug target hypotheses.
disable-model-invocation: true
---

## COMPUTE, DON'T DESCRIBE
When analysis requires computation (statistics, data processing, scoring, enrichment), write and run Python code via Bash. Don't describe what you would do — execute it and report actual results. Use ToolUniverse tools to retrieve data, then Python (pandas, scipy, statsmodels, matplotlib) to analyze it.

# Pathway-Disease Genetics: GWAS to Drug Targets via Pathways

Connect genome-wide association study (GWAS) variants to biological pathways for mechanistic understanding and drug target discovery.

## When to Use

- "What pathways are disrupted in [disease] based on GWAS?"
- "Which GWAS genes for [trait] are in druggable pathways?"
- "Map [SNP/variant] to its causal gene and pathway"
- "Find drug targets from GWAS data for [disease]"
- "What is the eQTL evidence for [gene] in [tissue]?"

---

## Core Reasoning Principles

### Gene-to-Pathway Reasoning

A gene found in GWAS doesn't tell you which pathway is dysregulated. To connect gene -> pathway -> disease mechanism, ask: what biological process does this gene participate in? Use Reactome/KEGG to find pathways, then ask: which of these pathways is relevant to the disease phenotype?

For example, TCF7L2 is the strongest T2D GWAS gene. It participates in the Wnt signaling pathway. The question is then: how does disrupted Wnt signaling impair beta-cell function or insulin secretion? That reasoning step — from pathway membership to disease mechanism — requires combining pathway data with tissue expression (GTEx) and disease biology.

Non-coding GWAS variants (the majority) rarely affect the nearest gene. They act through regulatory elements that alter expression of genes sometimes hundreds of kilobases away. Always check eQTL evidence before assuming the nearest gene is causal.

### Pathway Convergence

Multiple disease genes mapping to the same pathway is stronger evidence than a single gene. If 5 GWAS hits for a disease all map to the NF-kB pathway, that is strong mechanistic evidence — the pathway is likely causal, not just coincidentally hit. If GWAS genes scatter across unrelated pathways, the mechanism is unclear, and you may need to look at upstream regulators or gene network hubs that connect the scattered genes.

When running enrichment (Reactome, KEGG, STRING), prioritize pathways that appear across multiple databases. A pathway enriched in all three is more reliable than one that appears in only one analysis.

### Druggability Reasoning

A pathway with existing drugs targeting its components is more actionable than a novel pathway. Before proposing a target as novel, check: are any pathway members already drug targets? Use DGIdb and OpenTargets to survey approved and clinical-stage drugs in the pathway.

Priorities: (1) approved drug for a different indication hitting a GWAS-supported target = strong repurposing opportunity; (2) drug in clinical trials hitting a GWAS-supported target = accelerated validation path; (3) druggable gene with no existing drugs + strong GWAS evidence = novel target opportunity.

"Undruggable" by current modalities does not mean permanently undruggable. Flag such genes but do not dismiss them — they may be actionable via gene therapy, RNA therapeutics, or downstream pathway intervention.

---

## Tool Selection Guide

### Phase 1: Disease Resolution and GWAS Collection

Resolve disease name to ontology ID first:
- `OpenTargets_multi_entity_search_by_query_string(queryString=<disease>)` — returns EFO/MONDO IDs

Collect GWAS signals:
- `gwas_search_associations(query=<disease>)` — broad search
- `gwas_get_variants_for_trait(trait=<trait>, p_value_threshold=5e-8)` — genome-wide significant hits
- `gwas_get_snps_for_gene(gene_symbol=<gene>)` — gene-centric search

**Gotcha:** `gwas_get_associations_for_trait` is broken — use `gwas_search_associations` instead. `gwas_get_snps_for_gene` uses `gene_symbol`, not `mapped_gene`.

### Phase 2: Variant Annotation and eQTL Evidence

Annotate variants:
- `EnsemblVEP_annotate_rsid(variant_id=<rsid>)` — functional consequence, nearest gene
- Response format is variable: list, `{data, metadata}`, or `{error}` — handle all three

Query eQTL evidence in tissue relevant to the disease (e.g., pancreas for T2D, brain for neurological):
- `GTEx_query_eqtl(gene_input=<gene>, tissue=<tissue>)` — never pass empty `gene_input`
- `GTEx_get_expression_summary(gene_input=<gene>)` — expression across all tissues
- `GTEx_get_median_gene_expression(gencode_id=[<versioned_id>], tissue_site_detail_id=[<tissue>])` — use versioned Ensembl IDs (e.g., `ENSG00000148737.11`) and `gtex_v8`

### Phase 3: Pathway Enrichment

Run enrichment across multiple databases and cross-validate results:

- `ReactomeAnalysis_pathway_enrichment(identifiers="P04637 P38398 ...")` — space-separated UniProt STRING, not an array
- `Reactome_map_uniprot_to_pathways(uniprot_id=<id>)` — per-gene pathway membership
- `Reactome_get_participants(pathway_id=<R-HSA-XXXXX>)` — all genes in a pathway
- `KEGG_get_gene_pathways(gene_id=<kegg_id>)` — KEGG pathways for one gene
- `kegg_search_pathway(query=<disease_or_process>)` — keyword search
- `STRING_functional_enrichment(protein_ids=[<genes>], species=9606)` — GO/KEGG/Reactome with FDR
- `PANTHER_enrichment(gene_list="GENE1,GENE2,...", organism=9606, annotation_dataset="GO:0008150")` — comma-separated STRING, not array

**MetaCyc note:** Currently unavailable (BioCyc requires authentication). Use KEGG or Reactome for metabolic pathways.

### Phase 4: Druggability and Drug Landscape

- `DGIdb_get_gene_druggability(genes=[<gene_list>])` — categories: clinically actionable, druggable, etc.
- `DGIdb_get_drug_gene_interactions(genes=[<gene_list>])` — use `genes` param (array), not `gene_name`
- `OpenTargets_get_associated_drugs_by_target_ensemblID(ensemblId=<id>)` — approved and clinical drugs
- `OpenTargets_target_disease_evidence(ensemblId=<id>, efoId=<disease_id>)` — genetic + other evidence score

---

## Three-Step Workflow

### Step 1: GWAS to Causal Gene

1. Resolve disease ID via `OpenTargets_multi_entity_search_by_query_string`
2. Pull GWAS hits with `gwas_get_variants_for_trait` (p < 5e-8)
3. Annotate each lead SNP with VEP — is it coding or non-coding?
4. For non-coding variants, check eQTL via GTEx in the disease-relevant tissue
5. Prioritize genes where: GWAS SNP is also a significant eQTL AND the tissue is biologically relevant to the disease
6. Cross-check with `OpenTargets_target_disease_evidence` for additional genetic evidence

Evidence tiers: **High** = GWAS p < 5e-8 + eQTL colocalization in relevant tissue + coding variant; **Medium** = GWAS p < 5e-8 + eQTL in any tissue; **Low** = GWAS p < 5e-8 + positional mapping only.

### Step 2: Gene Set to Pathway Enrichment

1. Collect prioritized gene list from Step 1
2. Run `ReactomeAnalysis_pathway_enrichment` and `STRING_functional_enrichment`
3. Map each gene to KEGG pathways via `KEGG_get_gene_pathways`
4. Identify pathways appearing across multiple databases (convergence = stronger evidence)
5. For metabolic diseases, add tissue-specific network context via `humanbase_ppi_analysis` (all 5 params required: `gene_list`, `tissue`, `max_node`, `interaction`, `string_mode`)
6. Rank pathways by enrichment FDR x number of GWAS genes x biological plausibility

### Step 3: Pathway to Drug Target

1. From enriched pathways, extract all member genes via `Reactome_get_participants` and `KEGG_get_pathway_genes`
2. Assess druggability via `DGIdb_get_gene_druggability`
3. Look up existing drugs via `OpenTargets_get_associated_drugs_by_target_ensemblID`
4. Cross-reference pathway genes with GWAS genes: overlap = top candidate
5. Classify each candidate: repurposing opportunity / novel target / undruggable (for now)

Final ranking: Genetic Evidence x Druggability x Pathway Centrality. Flag novel targets (strong genetic + no existing drugs) and repurposing opportunities (approved drug + genetic support in this disease).

---

## Key Parameter Gotchas

- `gwas_get_snps_for_gene`: use `gene_symbol`, not `mapped_gene`
- `OpenTargets_multi_entity_search_by_query_string`: use `queryString`, not `query`
- `GTEx_query_eqtl`: `gene_input` must never be empty
- `GTEx_get_median_gene_expression`: use versioned gencode IDs; use `gtex_v8`
- `ReactomeAnalysis_pathway_enrichment`: `identifiers` is space-separated STRING, not array
- `DGIdb_get_drug_gene_interactions`: use `genes` (array), not `gene_name`
- `PANTHER_enrichment`: `gene_list` is comma-separated STRING, not array
- `humanbase_ppi_analysis`: all 5 params required
- `EnsemblVEP_annotate_rsid`: use `variant_id`, not `rsid`
- `kegg_find_genes`: include `organism="hsa"` for human genes

---

## Limitations

- GTEx eQTL lookup is not formal statistical colocalization (coloc/ENLOC) — treat as suggestive evidence
- GWAS Catalog may not include recent publications; cross-check with Open Targets
- Reactome and KEGG define pathways differently; some biology is in one but not the other
- DGIdb druggability categories are heuristic — "undruggable" applies only to current modalities
- eQTLs are tissue-specific; querying the wrong tissue may miss causal effects

---

## Related Skills

- **tooluniverse-gwas-trait-to-gene**: Focused GWAS-to-gene mapping
- **tooluniverse-gene-enrichment**: Detailed enrichment analysis
- **tooluniverse-drug-target-validation**: Deep target validation
- **tooluniverse-network-pharmacology**: Network-level drug analysis
- **tooluniverse-variant-functional-annotation**: Detailed variant interpretation

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

Similar skills

All skills for AI and ML →
scientific-thinking-biology Agents365-ai scientific-thinking-biology is a Claude Code skill that structures reasoning about biological research and genomic data. It distinguishes association from causation and marker from driver. CX Codex +1 296installs pathway-enrichment K-Dense-AI pathway-enrichment runs pathway and gene-set enrichment analysis on gene lists or ranked data, finds over-represented GO terms, KEGG, Reactome, MSigDB. Needed after differential expression, CRISPR screens, or clustering. CX Codex +1 1,600installs gtex-eqtl-skill official OpenAI The gtex-eqtl-skill skill fetches GTEx single-tissue eQTL associations for one variant and turns rsID, GRCh37, or GRCh38 input into a GRCh38 query. It returns JSON. CX Codex GP ChatGPT +1 13installs gtex-database google-deepmind gtex-database retrieves RNA expression and eQTL data from GTEx across 54 non-diseased human tissue sites. Use it to query genes, variants, or tissue-specific expression. CX Codex +1 2,300installs epigraphdb-skill official OpenAI This skill sends requests to EpiGraphDB and turns the responses into concise summaries. Use it to look up ontology, literature, Mendelian randomization, gene-drug, and support-path evidence. CX Codex GP ChatGPT +1 13installs open-targets-search yorkeccak open-targets-search — a skill that asks Open Targets about drug-disease associations in plain language, with Valyu handling the semantic search. Handy when you want target evidence without assembling API parameters. CX Codex +1 130installs

From the same set136

The whole set →
tooluniverse-structural-proteomics tooluniverse-structural-proteomics is a Claude Code skill that gathers structural and proteomics data to assess drug targets. It supports binding-site, ligand, and model-confidence analysis. CX Codex +1 240installs tooluniverse-small-molecule-discovery tooluniverse-small-molecule-discovery is a Claude Code skill for finding and analyzing small molecules, from compound names to structures, activity, ADMET properties, and availability. CX Codex +1 240installs tooluniverse-gene-disease-association tooluniverse-gene-disease-association is a skill that cross-references gene-disease associations across databases and scores evidence concordance for clinical and research queries. CX Codex +1 240installs tooluniverse-protein-modification-analysis tooluniverse-protein-modification-analysis is a Claude Code skill for studying protein post-translational modifications. It handles PTM site annotation, enzyme links, and disease associations. CX Codex +1 239installs tooluniverse-lipidomics tooluniverse-lipidomics is a Claude Code skill that identifies and classifies lipids, maps metabolic pathways, and retrieves disease associations for lipidomics analysis. CX Codex +1 238installs tooluniverse-comparative-genomics tooluniverse-comparative-genomics is a Claude Code skill that finds orthologs and assesses gene conservation across species. CX Codex +1 238installs tooluniverse-metabolomics-pathway The tooluniverse-metabolomics-pathway skill connects metabolites with pathways, enzymes, genes, and diseases. Use it when metabolomics work needs cross-database identification and mapping. CX Codex +1 238installs tooluniverse-electron-microscopy The tooluniverse-electron-microscopy skill for Claude Code searches and inspects electron microscopy records in EMDB, PDB, EMPIAR, and CryoET Data Portal. It helps link density maps to fitted atomic models. CX Codex +1 238installs

Project sponsors

Companies that keep this analytics open