tooluniverse-mendelian-randomization

Use Mendelian randomization, which estimates causal effects using genetic variants as instrumental variables. Resolve the exposure and outcome traits to exact OpenGWAS labels first, then query the pre-computed MR estimate and interpret the beta, p-value, and MOE instrument quality score.

Genetic correlation (rg) means two traits share heritability but says nothing about causal direction. Mendelian randomization uses genetic instruments to test whether one trait causally affects another, so only MR supports a causal claim.

EpiGraphDB matches trait labels exactly and case-sensitively, so free-text names often miss. Resolve the trait with the OpenGWAS search first and read the metadata note; an empty result usually means a label mismatch, not absence of a causal effect.

The pre-computed MR-EvE resource covers curated trait pairs only. For custom pairs or instrument thresholds, fetch harmonized exposure and outcome SNP data via OpenGWAS (requires a free OPENGWAS_JWT) and compute IVW or MR-Egger estimates yourself or with the TwoSampleMR R package.

MR rests on three assumptions: relevance, independence, and exclusion restriction. Horizontal pleiotropy is the dominant threat, most instruments are European-ancestry, and weak instruments bias toward the observational estimate, so report MOE scores, method agreement, and ancestry.