66. pleiotropic analysis of cognitive genomics uncovers shared and specific biology of schizophrenia and bipolar disorder

Background Bipolar disorder (BIP) and schizophrenia (SCZ) show high genetic correlation (Rg≈0.6–0.7), yet exhibit substantial heterogeneity in symptom profiles, disease progression, and treatment responses, both between and within these disorders. While genome-wide association studies (GWAS) have identified hundreds of risk loci, these findings often conflate distinct biological processes, obscuring the pathways driving shared versus disorder-specific risk. We have previously demonstrated that pleiotropy of SCZ with cognitive phenotypes (cognitive task performance and educational attainment) can be used to identify dissociable biological pathways in schizophrenia. In this report, we extend this approach to both the shared and disorder-specific components of SCZ and BIP. Methods Using Genomic SEM, we regressed SCZ and BIP onto three latent factors: a shared factor (PSYshared) and two disorder-specific factors (SCZ-specific, BIP-specific). We then used PLEIO to estimate joint effects of the cognitive phenotypes across each of the 3 latent psychiatric factors. As in our prior work (cited above), PLEIO output was then categorized into "Concordant and "Discordant" loci. Concordant loci are those in which the allele increasing risk for psychiatric disorder also is associated with cognitive impairment. By contrast, at Discordant loci, the risk allele is paradoxically associated with enhanced cognitive performance and/or educational attainment. Results Genomic SEM isolated a common transdiagnostic psychosis liability factor (PSYshared) and disease specific factors which showed near-zero correlation with the opposite disorder (SCZ-specific to BIP: rg = 0.098; BIP-specific to SCZ: rg = −0.117) demonstrating clean separation of shared vs. unique genetic risk. Results for PSYshared were similar to those observed in our prior work for schizophrenia alone: specifically, Concordant loci implicated core neurodevelopmental biological processes, while Discordant loci implicated synaptic pathways. New biology was uncovered in the pleiotropic analysis of cognition with the disorder-specific phenotypes. For the SCZ-Specific factor: Concordant loci were enriched for alternative splicing pathways and processes related to neuron projection/fate, while Discordant loci implicated histone modification processes. For the BIP-Specific factor: Concordant loci were enriched for synapse structure/assembly, transcriptional regulation, and alternative mRNA splicing. Discordant loci were enriched for metabolic pathways (specifically, catabolism). Discussion By decomposing SCZ and BIP into shared and disorder-specific genetic factors, and examining distinct patterns of pleiotropy with cognitive phenotypes, our analysis exposed biologically distinct pathway signatures that remain cryptically entangled in readouts from standard GWAS. In addition to providing biological insights into the similarities and differences between SCZ and BIP, these differential pathways can be used to form the basis of data-driven PRS subsets.