agentsclimarketplace

Causal evidence

Skill Alim430/bioresearch-agent/skills/biomedical/causal-evidence

Executable biomedical workflows for AI assistants — literature analysis, biomarker discovery, and Mendelian randomization through reproducible agent skills.

Install
npx -y skills add Alim430/bioresearch-agent --skill causal-evidence

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Run the causal-evidence chain (GWAS → eQTL → colocalization → TWAS → fine-mapping → MR) via the BioResearch Agent causal-evidence workflow. Use when the user wants to test whether a GWAS risk locus acts through a candidate gene's expression, identify the credible set, and check MR consistency. Synthetic loci by default; methodology validation, not a real etiological claim.

SKILL.md

2.2 KB, as published. Nobody here has run it

BioResearch Agent — Causal Evidence Chain Skill

Capability

Runs the full causal-evidence chain for a set of loci:

  1. GWAS → eQTL — per-SNP association in both trait and expression.
  2. Colocalization — full 5-hypothesis coloc (Giambartolomei 2014), reporting PP.H4 (shared causal variant) vs PP.H3 (distinct variants). Numerically stable (log-space ABF).
  3. TWAS — S-PrediXcan-style expression-trait association (Z, p).
  4. Fine-mapping — Bayesian credible set via per-SNP posterior inclusion probability (PIP), normalized over the locus.
  5. MR — Wald-ratio causal estimate for the colocalized gene's lead SNP.

Returns a per-gene evidence table + credible-set CSV + locus heatmap, not a biological claim.

Run

bioresearch run causal-evidence --seed 42 --output-dir outputs/causal-evidence

Outputs (in --output-dir)

  • CE_per_gene_results.csv — per-gene: truth class, PP.H4, TWAS Z/p, MR β/p, n credible
  • CE_credible_sets.csv — per-SNP PIP + credible-set membership
  • CE_locus_heatmap.png — GWAS / eQTL / coloc association heatmap
  • CE_recovery_benchmark.csv — ground-truth recovery across effect sizes
  • CE_summary_report.txt — human-readable summary
  • CE_evidence_package.json — reproducible Evidence Package (provenance + benchmark + grade)

Note

This skill dispatches to the framework's causal-evidence workflow / demo_causal_evidence.py. It adds no analysis of its own; all statistics run in the workflow modules. By default uses synthetic loci with ground-truth labels to validate the engine — real-data AD reference loci (TREM2 / BIN1 / APOE / CLU / PICALM) are listed as the target for a real-data version. Evidence grade is C (methodology validation).

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.