A single variant, whole-exome linear mixed model association analyses were performed with the REGENIE software package51 (link) v2.0.2 in two steps:
REGENIE Step 1: Whole-genome regression model using the recommended parameter setting. The UK Biobank DNA microarray genotypes (hap_v2) were lifted to build GRCh38 and filtered using the recommended settings (genotype call rate > 0.1; Hardy–Weinberg equilibrium p-value < 1e−15; MAC > 100; MAF > 0.01; sample call rate > 0.90).
REGENIE Step 2: Association analysis was performed on imputed dosage levels (BGEN v1.2 8-bit data format) for 408,511 individuals and 1,844,188 variants. LD was estimated afterward between all pairs of significantly associated variants using PLINK v1.90b252 (link). We restricted the association analysis to variants with MAC ≥ 5, as recommended by the REGENIE developers51 (link).
To calculate association statistics independent of common variants, we extracted 423, 63, and 114 independent, common SNPs previously reported to be associated with eGFR4 (link),28 (link), UACR9 (link), and urate18 (link), respectively, from the imputed UK Biobank data, set. We calculated residuals by regressing eGFR, UACR, and urate on all SNPs for a given trait in one regression analysis, and used these residuals as phenotypes in another ExWAS. Adjusted and unadjusted association results were compared with respect to their effect size estimates. Reported R2 measures originated from a linear regression model.
The PheWeb software (https://github.com/statgen/pheweb ) with default settings was used to create a local PheWeb instance that displays the results of the single variant analysis and is available under https://ckdgen-ukbb.gm.eurac.edu/ .
REGENIE Step 1: Whole-genome regression model using the recommended parameter setting. The UK Biobank DNA microarray genotypes (hap_v2) were lifted to build GRCh38 and filtered using the recommended settings (genotype call rate > 0.1; Hardy–Weinberg equilibrium p-value < 1e−15; MAC > 100; MAF > 0.01; sample call rate > 0.90).
REGENIE Step 2: Association analysis was performed on imputed dosage levels (BGEN v1.2 8-bit data format) for 408,511 individuals and 1,844,188 variants. LD was estimated afterward between all pairs of significantly associated variants using PLINK v1.90b252 (link). We restricted the association analysis to variants with MAC ≥ 5, as recommended by the REGENIE developers51 (link).
To calculate association statistics independent of common variants, we extracted 423, 63, and 114 independent, common SNPs previously reported to be associated with eGFR4 (link),28 (link), UACR9 (link), and urate18 (link), respectively, from the imputed UK Biobank data, set. We calculated residuals by regressing eGFR, UACR, and urate on all SNPs for a given trait in one regression analysis, and used these residuals as phenotypes in another ExWAS. Adjusted and unadjusted association results were compared with respect to their effect size estimates. Reported R2 measures originated from a linear regression model.
The PheWeb software (
Free full text: Click here
