Genomic Structural Equation Modelling of GWAS Summary Statistics Outline the Genetic Architecture of Anxiety Disorder

Student thesis: Master's ThesisMaster of Science


Background. Latent factor modelling using genetic data is one approach to disentangle variants with related and more specific biological functions and to find the different factors influencing anxiety and depression, increasing our understanding of the biology of anxiety symptoms. Aims/Hypotheses. My aim was to use genomic structural equation modelling (genomic SEM) to explore latent factor structures based on genomic associations with traits assessed by their genetic correlations with general anxiety disorder. Method. We used GWAS summary statistics from the UK Biobank on self-reported anxiety disorder by professional diagnosis or likely lifetime DSM-IV generalised anxiety disorder obtained through the UK Biobank Mental Health questionnaire ( N case =25,453; N control =58,113 ). Based on the genetic correlation with the anxiety disorder trait (|r g |>0.4, p<1e-5) computed by LD Score regression (LDSC), summary statistics for alcohol dependence ( N case =11,569; N control =34,999 ), self-rated health ( N=111,483 ), neuroticism ( N=170,911 ), subjective well-being ( N=298,420 ), self-reported tiredness ( N=108,976 ), and clinically ascertained major depressive disorder ( N case =16,823; N control =25,632 ) were included for analysis. An exploratory factor analysis was performed to inform confirmatory factor analyses of proposed latent factor genomic SEM models. Results. I found solid support for a genomic SEM two-factor model where anxiety disorder, major depressive disorder, neuroticism, and subjective well-being were characterised by a first latent factor (|std. pat. coeff.| >0.8) while self-rated health and self-reported tiredness loaded on a second factor (|std. pat. coeff.| >0.79), with a strong genetic correlation between the factors (r >0.71). Weak support was found for a one-factor model. Conclusions. Results suggest the presence of common but also distinct genetic effects influencing either internalising disorders or self-rated health and self-reported tiredness. We hypothesise that our factor models are mainly informed by trait-specific genetic effects but additional replication is required to confirm the results. Using an expanded set of traits may be desirable to improve the utility of similar genomic factor models and when generating latent factor GWASs that may be useful for biological pathway and other analyses.
Date of Award16 Oct 2020
Original languageEnglish
Awarding Institution
  • King's College London
  • Institute of Psychiatry, Psychology & Neuroscience
  • Social Genetic & Developmental Psychiatry
SupervisorGerome Breen (Supervisor) & Thalia Eley (Supervisor)

Cite this