Network analysis of the genetic relationships between twenty psychiatric and substance use disorders
TL;DR
Compared to standard factor analytic methods, network analyses can provide both confirmatory results and novel insights into the structure of genetic risks for a variety of psychiatric disorders.
Abstract
We sought to clarify, using network analysis, the structure of genetic risk factors for a wide range of psychiatric and substance use disorders. Using split halves (N=1,510,974), we computed the family genetic risk scores (FGRSs) for 20 psychiatric and substance use disorders assessed in treated individuals from the entire Swedish population born between 1960 and 2000. We calculated a genetic correlation matrix for them to which we applied an EBICglasso network analysis. The network was dense, with five inter-connected clusters of genetic risks for internalizing, externalizing, neurodevelopmental, eating, and psychotic disorders. Genetic risks for four disorders were transitional across clusters: attention-deficit/hyperactivity disorder (ADHD) between neurodevelopmental and externalizing disorders; bipolar disorder (BD) between internalizing and psychotic disorders; obsessive-compulsive disorder (OCD) between internalizing and eating disorders; and autism spectrum disorder (ASD) between neurodevelopmental and psychotic disorders. The most central nodes in the network were genetic risks for major depression (MD) and drug use disorder. The strongest links of genetic risk for BD were with genetic risks for MD and schizoaffective disorder, with no direct link with schizophrenia. We then added genetic risks for borderline personality disorder (BPD), suicide attempt (SA) and suicide death (SD). Genetic risk for BPD had diverse links to genetic risks for SA, BD, internalizing, externalizing, and eating disorders. Genetic risks for SA and SD had their strongest edges with, respectively, genetic risks for drug use and alcohol use disorder. Our results were stable across split halves and when sampling only one sibling per family. Compared to standard factor analytic methods, network analyses can provide both confirmatory results and novel insights into the structure of genetic risks for a variety of psychiatric disorders.