Identifying genetically driven clinical phenotypes using linear mixed models

Jonathan D. Mosley, John S. Witte, Emma K. Larkin, Lisa Bastarache, Christian M. Shaffer, Jason H. Karnes, C. Michael Stein, Elizabeth Phillips, Scott J. Hebbring, Murray H. Brilliant, John Mayer, Zhan Ye, Dan M. Roden, Joshua C. Denny

Research output: Contribution to journalArticlepeer-review

8 Scopus citations


We hypothesized that generalized linear mixed models (GLMMs), which estimate the additive genetic variance underlying phenotype variability, would facilitate rapid characterization of clinical phenotypes from an electronic health record. We evaluated 1,288 phenotypes in 29,349 subjects of European ancestry with single-nucleotide polymorphism (SNP) genotyping on the Illumina Exome Beadchip. We show that genetic liability estimates are primarily driven by SNPs identified by prior genome-wide association studies and SNPs within the human leukocyte antigen (HLA) region. We identify 44 (false discovery rate q<0.05) phenotypes associated with HLA SNP variation and show that hypothyroidism is genetically correlated with Type I diabetes (rG=0.31, s.e. 0.12, P=0.003). We also report novel SNP associations for hypothyroidism near HLA-DQA1/HLA-DQB1 at rs6906021 (combined odds ratio (OR)=1.2 (95% confidence interval (CI): 1.1-1.2), P=9.8 × 10-11) and for polymyalgia rheumatica near C6orf10 at rs6910071 (OR=1.5 (95% CI: 1.3-1.6), P=1.3 × 10-10). Phenome-wide application of GLMMs identifies phenotypes with important genetic drivers, and focusing on these phenotypes can identify novel genetic associations.

Original languageEnglish (US)
Article number11433
JournalNature communications
StatePublished - Apr 25 2016

ASJC Scopus subject areas

  • General Chemistry
  • General Biochemistry, Genetics and Molecular Biology
  • General Physics and Astronomy


Dive into the research topics of 'Identifying genetically driven clinical phenotypes using linear mixed models'. Together they form a unique fingerprint.

Cite this