56. PARSING ADHD GENETIC RISK IN THE CONTEXT OF THE EXTERNALIZING SPECTRUM

Background Individuals diagnosed with ADHD often develop other “externalizing” problems, such as conduct problems and substance use, which are thought to co-occur in part due to shared etiology. However, little work to date has distinguished the genetic architecture and associated neurobiology that is unique to ADHD from that which is shared with other forms of externalizing psychopathology. Here, we address this through a series of multivariate genomic and transcriptomic analyses, mapping disorder-specific and transdiagnostic liabilities to distinct biological pathways. We also introduce a new method to test whether genetic overlap with exogenous traits is disproportionately attributable to specific latent pathways. Methods Using the Genomic SEM framework, we modeled a latent factor of externalizing (EXT) and a residual variance component of ADHD (ADHD_r) in GWAS data from 1.9 million individuals. We used our novel approach to decompose the genetic correlations between ADHD and 53 psychiatric and cognitive phenotypes, formally testing whether these relationships were disproportionately attributable to ADHD-specific versus EXT-shared pathways. We also evaluated how these patterns differed as a function of early- versus late-onset ADHD. To interrogate the biology involved, we conducted multivariate GWAS and TWAS analyses of EXT and ADHD_r, followed by gene property and gene-set enrichment analyses that quantified the differential involvement of neural cells, processes, and tissues. Finally, we estimated the genomic relationships between regional cortical expansion (CE) and both ADHD_r and EXT. Results ADHD shared 48% of its genetic variance with EXT. Genetic correlations between ADHD and internalizing and psychotic disorders were disproportionately driven by EXT-linked architecture, while genetic correlations with compulsive disorders were largely attributable to ADHD-specific architecture. Collectively, multivariate GWAS and TWAS prioritized 30 genes uniquely associated with ADHD_r, including several that have not previously been linked to ADHD (e.g., RPLP2, KCTD16, and CPNE7). Cell-type enrichment analyses implicated dopaminergic neurons in disorder-specific risk for ADHD. Imaging genetic analyses revealed significant negative genetic correlations between ADHD_r and CE in visual regions. On a cortex-wide scale, topographical annotation indicated that associations between ADHD_r and CE aligned with multiple functional and structural gradients, including age-related changes in excitation-inhibition balance and higher-granularity cytoarchitecture. Discussion Our study suggests that approximately half of ADHD genetic liability is not captured by EXT. We demonstrate that decomposing ADHD's genetic relationships with other psychiatric disorders reveals markedly different contributions from shared and specific architectures. Further, our multivariate GWAS and TWAS results begin to elucidate the genes and pathways involved in disorder-specific liability for ADHD, including novel associations in genes involved in synaptic organization and dopaminergic pathways. Imaging genetic results link disorder-specific architecture to differences in excitation-inhibition circuitry and lower-order sensory processing regions. Distinguishing these components of the genetic architecture of ADHD improves our understanding of its etiology and may help identify novel targets for risk stratification and treatment.