A phenome-wide association and Mendelian Randomisation study of polygenic risk for depression in UK Biobank

Abstract

<jats:title>Abstract</jats:title><jats:p>Depression is the leading cause of worldwide disability but there remains considerable uncertainty regarding its neural and behavioural associations. Depression is known to be heritable with a polygenic architecture, and results from genome-wide associations studies are providing summary statistics with increasing polygenic signal that can be used to estimate genetic risk scores for prediction in independent samples. This provides a timely opportunity to identify traits that are associated with polygenic risk of depression in the large and consistently phenotyped UK Biobank sample. Using the Psychiatric Genomics Consortium (PGC), 23andMe and non-imaging UK Biobank datasets as reference samples, we estimated polygenic risk scores for depression (depression-PRS) in a discovery sample of 10,674 people and a replication sample of 11,214 people from the UK Biobank Imaging Study, testing for associations with 210 behavioural and 278 neuroimaging phenotypes. In the discovery sample, 93 traits were significantly associated with depression-PRS after multiple testing correction. Among these, 92 traits were in the same direction, and 69 were significant in the replication analysis. For imaging traits that replicated across samples, higher depression-PRS was associated with lower global white matter microstructure, association-fibre and thalamic-radiation microstructural integrity (absolute β: 0.023 to 0.040, P<jats:sub>FDR</jats:sub>: 0.045 to 3.92×10<jats:sup>-4</jats:sup>). Mendelian Randomisation analysis showed a causal effect of liability to depression on these structural brain measures (β: 0.125 to 0.707, p<jats:sub>FDR</jats:sub>&lt;0.048). Replicated behavioural traits that positively associated with depression-PRS included sleep problems, smoking status, measures of pain and stressful life experiences, and those negatively associated with depression-PRS included subjective ratings of physical health (absolute β: 0.014 to 0.180, P<jats:sub>FDR</jats:sub>: 0.046 to 8.54×10<jats:sup>-15</jats:sup>). Effect of depression PRS on mental health in the presence of reported childhood trauma, stressful life events and those living in more socially deprived areas showed increased variance explained by 1.42 – 4.08 times (p<jats:sub>FDR</jats:sub> for their interaction with depression-PRS: 0.049 to 0.003). Overall, the present study revealed replicable associations between depression-PRS and white matter microstructure that appeared to be a causal consequence of liability to depression. Analyses provided further evidence that greater effects of polygenic risk of depression are found in individuals exposed to risk-conferring environments.</jats:p>

Epigenetic prediction of major depressive disorder

“…In addition to these, PRS were associated with disease and mental health variables, such as MDD, number of depressive episodes and Mood Disorder Questionnaire score. The results indicate that PRS were associated with variables relating to MDD manifestation, as shown in previous studies [ 16 ]. Moreover, although both risk scores were associated with sociodemographic measures, such as years of education and deprivation ranks, MRS had a stronger association than PRS, indicating a stronger role played by the MRS in environmental factors [ 8 – 12 ].…”

“…Recent phenome-wide association studies have shown that MDD PRS are associated with a range of psychosocial and mental health phenotypes [ 16 ]. Using the same approach, we tested the association between 61 behavioural phenotypes and compared the pattern of association between MRS and PRS.…”

“…Finally, to explore whether the MDD MRS and MRS-ns capture exposure to lifestyle factors associated with MDD, we tested the association between MDD MRS and MRS-ns, and 61 behavioural phenotypes and lifestyle factors. We compared these associations with those observed for PRS that have previously shown association with a wide range of neuropsychiatric traits [ 16 ].…”

Association Study of Genetic Variants in Calcium Signaling-Related Genes With Cardiovascular Diseases

“…Moreover, excluding participants with phecodes/diseases related to the investigating phecode/disease in the analysis prevented the contamination of the control population by cases, leading to an increased statistical power ( Wei et al, 2017 ). Our genetic variant-based PheWAS is also advantageous because the genotype is fixed at birth and the results are less susceptible to reverse causality ( Shen et al, 2020 ). The present study also has several limitations.…”

DNA methylation signatures of C-reactive protein associations with structural neuroimaging measures and major depressive disorder

“…All analyses were conducted using R (version 3.2.3) in a Linux environment. Linear mixedeffects models (function 'lme' in R package 'nlme') and general linear models (function 'glm' in R package 'stats') were used to investigate structural brain metrics 46,47 . False Discovery Rate (FDR) multiple comparison correction was applied to all bilateral/unilateral structures, lobes and white matter tracts, referred to as PFDR in this report, using the 'p.adjust' function in R and all betas were standardised.…”