Abstract
Despite notable progress in psychiatric genomics, there are no validated blood-based biomarkers for psychosis. Previous studies have failed to establish a link between schizophrenia polygenic scores (PGS) and blood protein levels. We aimed to identify associations between schizophrenia PGS and blood-based proteins, and to determine whether levels of 2077 proteins differ in individuals with psychosis. We analysed proteomic and genomic data from 47,969 participants in the UK Biobank. Association analyses in the 47,678 participants without psychosis (mean age 57.1 years, standard deviation 8.1 years; 54% female) identified nominal associations (p < 0.05) of schizophrenia PGS with 102 proteins. Four of these (TMPRSS15, ADGRB3, CEACAM21, and KLK1) met the false discovery rate (FDR) threshold of < 0.05. We investigated the association of these four proteins with psychosis in a matched case-control sample (283 cases, 849 controls, mean age 56.9 years, standard deviation 8.4 years; 48% female). In individuals with psychosis, we observed significantly lower levels of KLK1, even after adjusting for potential confounders (effect size -0.25, SE 0.09, FDR 0.049). This direction of effect was opposite to that observed in the primary analysis of individuals without psychosis (effect size 324.67, SE 48.32, FDR 3.85 × 10(-8)). The effect of antipsychotic medication did not explain this difference. This protein should be taken forward for further study and validation to investigate its potential as a psychosis biomarker.