Bioresearch Ancestry Aware Mr

BioResearch Agent — Ancestry-Aware MR Skill

Capability

Runs cross-ancestry Mendelian randomization with pleiotropy-aware methods:

1. Per-ancestry IVW — inverse-variance-weighted MR per ancestry with Cochran's Q heterogeneity

test, producing ancestry-specific causal effect estimates and confidence intervals.

1. Cross-ancestry meta-analysis — fixed-effects (FE) and random-effects (RE, DerSimonian-Laird)

meta-analysis across ancestries, with Cochran's Q and I² statistics to quantify cross-ancestry heterogeneity.

1. CAUSE-like model — (Morrison et al. 2020) EM algorithm modeling both correlated and

uncorrelated horizontal pleiotropy. Tests H0 (no causal effect, pleiotropy free to vary) vs H1 (causal effect + pleiotropy) via likelihood-ratio test. Distinguishes true causation from correlated pleiotropy — the key confound in standard MR.

1. MRMix-like model — (Wang et al. 2020) three-component mixture model (causal / pleiotropic /

null) via EM, estimating the proportion of pleiotropic instruments and adjusting the causal estimate accordingly.

1. Portability assessment — evaluates whether the causal effect estimated in a reference ancestry

(e.g., EUR) transfers to others: direction consistency, significance consistency, heterogeneity (I²), EUR-centric bias, and a composite transferability score (0–1).

Returns per-ancestry MR + meta-analysis + CAUSE/MRMix + portability report, not a definitive causal claim.

Run

bioresearch run ancestry-aware-mr --n-snps 200 --n-instruments 40 --true-effect 0.30 --seed 42 --output-dir outputs/ancestry-mr

Outputs (in --output-dir)

• amr_per_ancestry_results.csv — per-ancestry IVW results (ancestry, beta, se, CI, p, Q, p_het)
• amr_cross_ancestry_meta.csv — FE + RE meta-analysis (method, beta, se, CI, p, Q, I², tau²)
• amr_cause_results.csv — CAUSE-like model per ancestry (theta, eta, sigma, LRTstat, pvalue)
• amr_mrmix_results.csv — MRMix-like model per ancestry (theta, picausal, pipleiotropic, pi_null)
• amr_portability.csv — portability assessment (directionconsistency, I², transferabilityscore)
• amr_forest_plot.png — forest plot of per-ancestry + meta-analysis causal estimates
• amr_evidence_package.json — reproducible Evidence Package (provenance + parameters + grade)

Note

This skill dispatches to the framework's ancestry-aware-mr workflow / demo_ancestry_aware_mr.py. It adds no analysis of its own; all computations run in the workflow modules. By default uses simulated multi-ancestry GWAS with known causal effect and controlled pleiotropy rates to validate the MR + pleiotropy detection pipeline — real-data deployment would use BBJ (EAS), FinnGen (EUR), TPMI (SAS), and All of Us (AMR/AFR) summary statistics with ancestry-matched LD panels. Evidence grade is C (methodology validation). Part of Phase 3a (cross-ancestry MR, CPU-only).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Alim430
• Source: Alim430/bioresearch-agent
• License: MIT

Install and usage instructions live in the source repository linked above.