Abstract
Genomic structural equation modeling (GenomicSEM) has proven successful to unravel genetic loci associated with latent traits underlying multiple biomarkers. To identify kidney-specific genetic variants, we applied GenomicSEM to European-ancestry genome-wide association study summary statistics of four kidney function traits released by the CKDGen Consortium and UK Biobank (n≥343,836).
Methods: We processed ~6 million genetic variants across traits (minor allele frequency≥0.005; imputation quality score≥0.6), estimating pairwise genetic correlations via linkage disequilibrium-score regression. Using the “commonfactorGWAS” function in “GenomicSEM” v0.0.5c, we identified one latent factor (F_kidney) and estimated genetic associations via weighted least squares. Loci were functionally characterized using FUMA v1.5.0.
Results: We identified 195 loci (+/-250kb around the most associated variant) encompassing 1,108 independent variants (LD r2<0.6; P<5.0×10-8) associated with F_kidney: 5 new loci were significantly associated with F_kidney but not with any single biomarker and 39 were significantly associated with all four biomarkers. Gene-set analysis using MAGMA v1.07 showed higher tissue-specific enrichment for genes associated with F_kidney compared to single-biomarker analyses, in the kidney cortex (P=2.2×10-19) and medulla (P=5.7×10-19).
Conclusion: GenomicSEM of multiple kidney traits can help underpin genetic architecture of the unobservable kidney function. Further integration of kidney-specific tissues and deeper biological annotations are warranted to identify relevant molecular targets.