All genome-wide association (GWA) analyses were conducted separately on sample 1 and sample 2. GWA analyses of the 12 neuroticism items were conducted using logistic regression as implemented in PLINK 1.937,38, regressing the dichotomous neuroticism item responses on the imputed hard-called SNPs. Linear regression was used to analyze the neuroticism sum-score and both item cluster scores19. Sex, age, and townsend deprivation index (TDI; a measure based on postal code indicating material deprivation), were included in the analyses as covariates. Genotype array was only included as covariate in the analyses of sample 1 as the same array was used for all subjects in sample 2. Additionally, we included the first 10 genetic PCs as covariates to control for potential population stratification. Genetic PCs were computed separately for both samples using FlashPCA 239 on individuals of European ancestry, after LD pruning and filtering out SNPs with MAF < 0.01 and genotype missingness > 0.05.
