Abstract / Summary
Understanding the genetic architecture of schizophrenia (SCZ) is invaluable for the development of personalized treatment. In three independent cohorts, using an automated pipeline, we calculated 413 psychiatry-relevant polygenic scores. Using these scores, machine learning was applied to stratify SCZ patients into two genetic risk profiles (GRP) in FinnGen (nSCZ=7 486), and to validate these results in the PsyCourse Study (nSCZ=419) and Bari (nSCZ=531). While the two GRPs showed comparable polygenic SCZ risk, they were primarily distinguished by a greater predisposition for neuroticism, depression-related traits, and lower cognitive performance in GRP 1. The genetic prediction of GRP 1 was phenotypically characterized by an increased prevalence of SCZ and a more severe and complex clinical manifestation. This illustrates that the penetrance of genetic SCZ risk might partially depend on the predisposition for the aforementioned traits through pleiotropic variants. Our results provide novel, replicable insights into the genetic architecture of SCZ and might inform future personalized treatments.