Network-wide risk convergence in gene co-expression identifies reproducible genetic hubs of schizophrenia risk.

The omnigenic model posits that genetic risk for traits with complex heritability involves cumulative effects of peripheral genes on mechanistic “core genes,” suggesting that in a network of genes, those closer to clusters including core genes should have higher GWAS signals. In gene co-expression networks, we confirmed that GWAS signals accumulate in genes more connected to risk-enriched gene clusters, highlighting across-network risk convergence. This was strongest in adult psychiatric disorders, especially schizophrenia (SCZ), spanning 70% of network genes, suggestive of super-polygenic architecture. In snRNA-seq cell type networks, SCZ risk convergence was strongest in L2/L3 excitatory neurons. We prioritized genes most connected to SCZ-GWAS genes, which showed robust association to a CRISPRa measure of PGC3 regulation and were consistently identified across several brain regions. Several genes, including dopamine-associated ones, were prioritized specifically in the striatum. This strategy thus retrieves current drug targets and can be used to prioritize other potential drug targets. •We test risk gradients in gene co-expression networks based on the omnigenic model•Across psychiatric disorders, schizophrenia risk best fits the model’s predictions•Findings hold in networks from excitatory neurons and in iPSC-derived neurons•Prioritizing genes most connected to schizophrenia risk reveals potential targets Genetic risk for schizophrenia spans the entire genome. Borcuk et al. show that network neighbors of risk genes also harbor risk, most evidently in excitatory neurons. They leverage this property to propose potential risk genes that are “guilty by association.” Stem cell data support the link between these genes and schizophrenia risk.