Abstract
Background: Chronic kidney disease affects >10% population worldwide and ranks among the top causes of death. Due to disease complexity and lack of specific biomarkers, genome-wide structural equation modeling (GenomicSEM) has emerged as a potential method for identifying molecular targets of unobservable traits that reflect a series of observed biomarkers. We applied GenomicSEM to European-ancestry GWAS summary statistics released by the CKDGen Consortium and by UK Biobank, identifying a single latent kidney function factor in which we observed enhanced detection of missense variants, compared to the underlying GWAS traits.
Material and Methods: Using approximately 6 million variants from GWAS summary statistics for eGFRcrea and eGFRcys, BUN, and UA, we employed the "commonfactorGWAS" function from GenomicSEM v0.05c to estimate the GWAS of the corresponding latent kidney function trait. Genome-wide significant signals from all the traits were identified and tested for independence using GCTA-COJO for multi-SNP-based
joint and conditional analysis. Independent signals were annotated using VEP.
Results: Based on Fisher's exact test, we observed a significant increase in the proportion of missense variants among of the generated credible sets in the latent kidney function trait compared to standard traits such as eGFRcrea (P-value=0.0004998) eGFRcys (P-value=0.0009995).
Conclusion: Application of genome-wide SEM analysis to kidney function traits enhances the detection of functional variants for further molecular target prioritization.