In the current study, we sought to compare these strategies by conducting a suite of genetic analyses for depression and stratified subtypes in two UK-based cohorts: GS:SFHS11,12 and UK Biobank (UKB)13,14. To maximize the sample size, an unstratified analysis was initially conducted. This used MDD diagnostic information obtained at structured clinical interview15 in GS:SFHS (2603 cases, 16,122 controls), and a probable MDD phenotype obtained from a touchscreen questionnaire16, previously validated by Smith et al17, in UKB (8248 cases, 16,089 controls). Subsequent analyses stratified the phenotype on the basis of recurrence or sex. Each approach was evaluated using several metrics: the successful identification of variants reaching genome-wide significance in GWAS meta-analysis, an increased SNP-based heritability estimate, identification of significant genetic correlations with other traits using LD score regression, and increased variance explained by polygenic profile scores for MDD derived from three independent cohorts. These metrics aim to test whether basic stratification of the MDD phenotype improves etiological insight.
