Mendelian Randomization

This page is a placeholder for the detailed Mendelian randomization guide. MR is included in the association guide because it answers a causal-inference question that is different from ordinary observational association.

Question

Use MR when the question is:

Does genetic-instrument evidence support a causal effect of exposure X on outcome Y?

MR uses genetic variants as instruments for an exposure. Its assumptions and diagnostics are separate from ordinary regression adjustment.

Intended Pattern

mr <- associate$mr(
  data,
  exposure = exposure_trait,
  outcome = outcome_trait,
  instruments = variants,
  method = c("ivw", "egger", "median")
)

The exact supported instrument sources and estimator set are documented on the reference page as the backend develops.

What MR Is Not

Question Workflow
Is PRS associated with dementia? associate$regression()
Does PRS association differ by sex? associate$compare(type = "heterogeneity")
Is the PRS effect mediated through pathology? associate$mediation()
Does genetic evidence support exposure causality? associate$mr()

Returned Object

MR results should return a PolyGeniusAssociations object with MR-specific summary rows.

Schema

Column Meaning
outcome, exposure Outcome trait and instrumented exposure trait.
estimator MR estimator, such as IVW, Egger, or median.
estimate, se, lower, upper Estimator-level causal-effect estimate and uncertainty.
statistic, p.value, adj.p.value Inference for the estimator row.
fit.id Artifact/diagnostic key.

Added Artifacts

MR artifacts are reserved for instrument summaries, harmonization details, leave-one-out results, and estimator diagnostics such as pleiotropy or weak instrument checks.

Plotting

MR outputs are naturally shown as:

  • estimator-level forest plots;
  • leave-one-out sensitivity views;
  • instrument-effect scatter plots;
  • diagnostic summaries for pleiotropy and weak instruments.

The detailed plotting guidance will be added with the full MR workflow.

Status

The MR guide is under construction. The page is present so users can see where MR fits in the association question map.