Splitting schizophrenia: divergent cognitive and educational outcomes revealed by genomic structural equation modelling

The genetic relationship between schizophrenia, IQ, and educational attainment (EA) is complex. Schizophrenia polygenic scores (PGS) are linked to lower IQ, whilst higher IQ-PGS correlates with reduced schizophrenia risk. Paradoxically, genetic predisposition to higher EA has been associated with increased schizophrenia risk, a relationship potentially confounded by genetic overlap between schizophrenia and bipolar disorder. Using a latent-variable Genomic Structural Equation Modelling approach to GWAS summary statistics for schizophrenia and bipolar disorder, we dissected the genetic contribution to schizophrenia, identifying 63 SNPs specifically associated with schizophrenia (SZspecific) and 78 shared with bipolar disorder (PSYshared). Both schizophrenia (rg = -0.22) and SZspecific (rg = -0.24) were genetically negatively correlated with IQ; correlations between bipolar disorder and PSYshared with IQ were less pronounced (both rg = -0.07). Schizophrenia exhibited no correlation with EA, yet the latent variables demonstrated divergent relationships; PSYshared was positively correlated (rg = 0.11), whereas SZspecific was negatively correlated (rg = -0.06). PGS analyses in the UK Biobank (n = 381,688), corroborated these divergent relationships, SZspecific-PGS was negatively associated with EA (β = -0.13, p < 2e-16), whereas the PSYshared-PGS was positively associated (β = 0.14, p < 2e-16). Mendelian randomisation provided additional support but also confirmed the presence of genetic pleiotropy. These findings underscore the utility of genetic methods in dissecting the heterogeneity of neuropsychiatric disorders, supporting the existence of two possible pathways to schizophrenia: one shared with bipolar disorder and another with greater neurocognitive impact.