domains / families / science / genetic-correlation

genetic-correlation

separate

Estimates bivariate genetic correlation (rg) between traits from GWAS summary statistics or individual-level genotypes using cross-trait LDSC, HDL, LAVA, rho-HESS, GREML-bivariate, Popcorn, and HDL-L. Use when quantifying shared genetic architecture between two traits, screening MR validity before causal inference, distinguishing global from locus-level rg, estimating trans-ancestry rg, separating partial from full causation via LCV gcp, or producing a STROBE-MR-compliant cross-trait sensitivity battery. Cross-trait LDSC intercept absorbs sample overlap and is NOT a bias; HDL is biased under sample overlap above ~5%. High rg between exposure and outcome motivates CHP-aware MR sensitivity (CAUSE, LHC-MR).

When to use

• quantifying shared genetic architecture between two traits
• screening MR validity before causal inference
• distinguishing global from locus-level rg
• estimating trans-ancestry rg
• separating partial from full causation via LCV gcp
• producing a STROBE-MR-compliant cross-trait sensitivity battery

When not to use

• tasks where sample overlap is above ~5% for HDL
• non-EUR sumstats with EUR LD scores for cross-trait LDSC
• loci with too few SNPs for LAVA or rho-HESS

Skill page

Source