Genome-wide association study of delay discounting in 134,935 23 and me participants

Delay discounting (DD) is a heritable transdiagnostic trait, or endophenotype, that has been implicated in multiple psychiatric diseases, including substance use disorders and attention deficit hyperactivity disorder. A prior genome-wide association study (GWAS) of DD identified genetic correlations with these and other traits but was underpowered for genome-wide discovery. In collaboration with 23 and Me, Inc., we collected responses to a 30-item delay discounting questionnaire from 134,945 research participants, and performed a GWAS assuming an additive genetic model that included age, sex, the first five genetic principal components, and indicator variables for genotype platforms as covariates. We further explored the genetic architecture of DD and the pleiotropic mechanisms with other outcomes using an array of genomic tools, such as MAGMA, H-MAGMA, S-MultiXcan and S-PrediXcan, LDSC and LAVA (Local Analysis of [co]Variant Annotation). We identified 14 significant loci associated with DD, with an estimated SNP-heritability of 9.86% (± 0.57%). Most of these loci (e.g., rs34645063, chr6q16.1, p=3.20E-13; rs3020805, chr16p11.2, p=6.50E-10) have been previously associated with various other behavioral traits, including risk-taking, alcohol consumption, educational variables and cognitive ability, and psychiatric disorders, as well as obesity and BMI. Genetic correlation analyses revealed significant associations with 27 traits, such as educational variables (e.g., years of education rg=-0.57, SE=0.03; intelligence rg=-0.39, SE=0.03), smoking behaviors (e.g., smoking initiation rg=0.32, SE=0.02; tobacco use disorder (TUD) rg=0.33, SE=0.03), risky behaviors (e.g., externalizing rg=0.30, SE=0.03), and health-related outcomes (BMI rg=0.28, SE=0.03). Local genetic correlation analysis revealed 20 significant bivariate loci between DD and these 27 other traits. Among them, the locus comprising the NCAM1-TTC12-ANKK1-DRD2 gene cluster was positively correlated with 12 traits, including substance use traits (i.e., drinks per week rg = 0.61; problematic alcohol use rg = 0.47; cannabis initiation rg = 0.63; TUD rg = 0.46; smoking initiation rg = 0.31; cigarettes per day rg = 0.34), and psychiatric disorders (schizophrenia rg = 0.52). Polygenic analyses in a hospital-based cohort (BioVU, N=69,447) showed that the DD polygenic risk score (PGS) was significantly associated with 127 medical traits across 16 categories, the strongest association being with TUD (p=1.21E-16) and mood disorders (p=5.26E-10). Beyond psychiatric disorders, the PGS for DD was also positively associated with medical phenotypes, such as Diabetes Mellitus (p=1.75e-06), ischemic heart disease (p=8.51e-06), and hypertension (p=1.63e-05). Our results support the polygenic architecture of DD, identifying novel significant loci and highlighting common genetic factors between DD and other psychiatric and somatic health outcomes. This work further establishes DD as a valuable endophenotype.