0033 Genome-wide association analysis of composite sleep scores in 413,904 individuals

Abstract

Abstract Introduction Recent genome-wide association studies (GWAS) of individual sleep traits have identified hundreds of genetic loci, often implicating genes and biological pathways that are shared across sleep phenotypes. Modeling multiple dimensions of sleep may allow identification of common underlying genetic mechanisms and complement analyses of individual traits. Moreover, construction of novel sleep health scores, accounting for correlations between traits, has the potential to enhance specificity and power for genetic analyses. Methods We performed GWAS of composite sleep scores derived from five sleep questions characterizing 413,904 individuals of European ancestry from the UK Biobank. We constructed an additive sleep score (SS-add) as a sum of up to five favorable self-reported sleep behaviors (sleep duration of 7-8 hours, early chronotype, few insomnia symptoms, no snoring, and no excessive daytime sleepiness) as well as five principal component scores (SS-PC1 – SS-PC5) using the underlying sleep traits. SNP-level association studies of the sleep scores were complemented with multiple follow-up analyses, including investigation of pathway, tissue, and cell-type enrichments, as well as comprehensive genetic correlation (381 representative phenotypes) and bi-directional Mendelian randomization (MR) analyses (40 selected phenotypes). Results Sleep scores SS-PC1 (interpretable as longer duration sleep without insomnia symptoms) and SS-PC2 (healthy sleep without snoring or sleepiness) showed higher heritability (respectively 11.7% and 9.3%) compared with their primary underlying traits. SS-PC3 (loading strongly on chronotype) showed highest overall heritability (15.3%). Accounting for the six GWAS, we identified 28 significant novel loci (p< 8.3e-9), 31 additional novel loci (p< 5e-8), and 341 loci previously reported (p< 5e-8) by GWAS of individual sleep traits. Associated loci mapped to genes enriched in expression in brain tissues, and in metabolic and neuronal pathways. Numerous neurological, cardiometabolic and other traits were genetically correlated with sleep health (104 with|rho_g|>0.3, p< 2e-5). MR associations (p< 2e-4) pointed to causes (BMI [SS-PC2/PC4], smoking [SS-add/PC1], lower socio-economic status [SS-add/PC5]) and consequences (coronary heart disease [SS-add/PC1], pain [SS-PC1/5]) of poor sleep. Conclusion Composite sleep health scores revealed novel genetic mechanisms of related sleep behaviors, helped clarify relationships with neurological and cardiometabolic traits, and deserve further investigation. Support (if any) R01HL153814 (to H.W.), NHLBI R35HL135818 (to S.R.), 1R01HL146751 (to R.S.), UKB Application 6818.