AbstractBackgroundSuicide is the leading cause of death in youth worldwide.1 Identifying children with high risk for suicide remains challenging.2 Here we test the extents to which genome-wide polygenic scores (GPS) for common traits and psychiatric disorders are linked to the risk for suicide in young children.MethodsWe constructed GPSs of 24 traits and psychiatric disorders broadly related to suicidality from 8,212 US children with ages of 9 to 10 from the Adolescent Brain Cognitive Development study. We performed multiple logistic regression to test the association between childhood suicidality, defined as suicidal ideation or suicidal attempt, and the GPSs. Machine learning techniques were used to test the predictive utility of the GPSs and other phenotypic outcomes on suicide and suicidal behaviors in the youth.OutcomesWe identified three GPSs significantly associated with childhood suicidality: Attention deficit hyperactivity disorder (ADHD) (P = 2.83×10−4; odds ratio (OR) = 1.12, FDR correction), general happiness with belief that own life is meaningful (P = 1.30×10−3; OR = 0.89) and autism spectrum disorder (ASD) (P = 1.81×10−3; OR = 1.14). Furthermore, the ASD GPS showed significant interaction with ELS such that a greater polygenic score in the presence of a greater ELS has even greater likelihood of suicidality (with active suicidal ideation, P = 1.39×10−2, OR = 1.11). In machine learning predictions, the cross validated and optimized model showed an ROC-AUC of 0.72 and accuracy of 0.756 for the hold-out set of overall suicidal ideation prediction, and showed an ROC-AUC of 0.765 and accuracy of 0.750 for the hold-out set of suicidal attempts.InterpretationOur results show that childhood suicidality is linked to the GPSs for psychiatric disorders, ADHD and ASD, and for a common trait, general happiness, respectively; and that GPSs for ASD and insomnia, respectively, have synergistic effects on suicidality via an interaction with early life stress. By providing the quantitative account of the polygenic and environmental factors of childhood suicidality in a large, representative population, this study shows the potential utility of the GPS in investigation of childhood suicidality for early screening, intervention, and prevention.
Is there any incremental benefit to conducting neuroimaging and neurocognitive assessments in the diagnosis of ADHD in young children? A machine learning investigation
