← ClaudeAtlas

bioresearch-causal-inferencelisted

Run two-sample Mendelian randomization via the BioResearch Agent causal workflow (IVW estimation, leave-one-out sensitivity, scatter / funnel plots). Use when the user asks to test a causal effect between an exposure and an outcome (e.g., BMI → Type 2 Diabetes) using genetic instruments.
Alim430/bioresearch-agent · ★ 1 · AI & Automation · score 72
Install: claude install-skill Alim430/bioresearch-agent
# BioResearch Agent — Causal Inference (MR) Skill ## Capability Runs a reproducible two-sample Mendelian randomization: GWAS summary statistics (simulated by default) → genome-wide significant SNP instruments → IVW causal estimate → leave-one-out sensitivity → scatter / funnel plots. Returns a quantitative causal estimate, not a claim. ## Run ```bash bioresearch run causal --exposure BMI --outcome "Type 2 Diabetes" ``` ## Outputs (in `outputs/causal/`) - `causal_ivw_results.csv` — IVW estimate, SE, p-value - `causal_loo_results.csv` — leave-one-out sensitivity - `causal_mr_scatter.png` — MR scatter plot (IVW slope labeled) - `causal_mr_funnel.png` — funnel plot - `causal_interpretation.txt` — interpretation notes ## Note This skill dispatches to the framework's `causal` workflow. It adds **no analysis of its own**; all MR statistics run in the workflow modules. By default uses simulated GWAS data (no external credentials required).