Identifying genomic and phenotypic risks factors for the clinical progression of depressive symptoms

Abstract

BACKGROUND: Major Depressive Disorder (MDD) is a leading cause of disability worldwide. It is a heritable condition with data from twin studies estimating a heritability of 37%. As sample sizes increase, Genome-wide association studies (GWAS) are revealing more genetic loci associated with MDD, however SNPs detected by GWAS explain only a fraction of the total heritability of MDD. This missing heritability could be due partially to available sample sizes being too small to detect rare variants and due to variation in how MDD is measured across cohorts with many phenotypic definitions of MDD existing. Furthermore, MDD is now thought to be a highly heterogenous disorder with individuals differing in their age of onset, symptomology, aetiology, and time course such as number and length of episodes and more. Until recently the genetic architecture and genetic loci associated with MDD subtypes have rarely been studied. A recent study carried out 16 subtype specific GWAS, with MDD subtypes showing differences in SNP heritability and genetic correlations between MDD subtypes and other related traits. Currently MDD status is often derived from questionnaires using self-reported and/or minimal phenotypes. Electronic health records including hospital and GP records could provide a useful more reliable alternative with MDD being reported by a professional as appose to the individual themselves.

THE OVERALL AIM of this thesis is to stratify MDD by its clinical severity and longitudinal course to identify risk factors and genetic loci using questionnaire, self-declared and electronic health record data from the UK Biobank and Airwave Health Monitoring Study and to consider how to define MDD phenotypically within these cohort studies.

CHAPTER 1 descries the study samples and gives a brief overview of genetic research to date. The datasets used in this thesis are the UK Biobank and the Airwave Health Monitoring Study. The UK Biobank cohort is a population-based longitudinal study of ~500,000 individuals, recruited at 23 centres across the UK, whose genotypic and environmental data were collected between 2006 and 2019 with participants being contacted up to 5 times. The Airwave Health Monitoring Study was initially established to evaluate possible health risks of using a TETRA communication system to members of the police force but has since been expanded to explore the general health of the police force including mental health. The cohort includes extensive phenotypic and genotypic data collection and will include longitudinal data collection. The study consists of ~50,000 individuals, of these currently ~20,000 participants have genotyped data

CHAPTER 2 aims to investigate whether genetic risk for MDD, as captured by a PRS, is associated with case-control status in a longitudinal population-based cohort. Firstly, the study examined how genetic risk of MDD was associated with MDD case-control status when using cross sectional or longitudinal data. Secondly the study examined if genetic risk of MDD was associated with number of times an individual was identified as a MDD case. Thirdly the study examined how participants’ genetic risk of MDD differs between item non-responders and those identified as controls and cases.

CHAPTER 3 aims to carry out a GWAS of depression using in a new cohort, Airwave Health Monitoring Study, for inclusion in the largest GWAS metanalysis of MDD to date and carry out genetic correlations with other phenotypes and other MDD cohorts in the metanalysis. Case-control status was determined using self-reporting a depression diagnosis. The Airwave cohort was found to have a lower genetic correlation with depression than other cohorts in the genome wide meta-analysis and so I investigated if refining the case-control status of individuals using both self-reported depression and PHQ9 scores improved the accuracy of case-control classification, quantified using MDD PRS and carrying out GWAS of the refined phenotype to determine if the genetic architecture of the refined phenotypes differs to self-reported depression.

CHAPTER 4 aims to use electronic health records to stratify MDD and carry out genome-wide association analyses of MDD subtypes, estimation of the heritability of MDD subtypes from GWAS data and identification of phenotypic and genetic correlations with other relevant phenotypes.

IN CONCLUSION, this thesis reports misclassification of controls and phenotypic heterogeneity can impact the estimates of genetic components. Using longitudinal data has the potential to increase power by reducing the proportion of misclassified cases and stratification of MDD by clinical progression may identify more homogenous subtypes of depression. These findings are subject to a number of strengths and limitations, which are discussed in the thesis along with suggestions for future work.