Plasma proteome profiling identifies novel biomarkers and predictors for schizophrenia

This study performed a systematic investigation through proteomic and longitudinal follow-up data from the UK Biobank (comprising 70 schizophrenia cases and 36,135 healthy controls) to reveal the association between protein biomarkers and the onset of schizophrenia in order to provide insights into the pathogenesis of schizophrenia. Cox regressions recognized 31 significant schizophrenia related proteins (P < 1.712 × 10^-5) and five of them also had non-linear associations with schizophrenia. The joint effect of 31 proteins was measured through Lasso-Cox regression, in which 17 proteins had non-zero coefficients. Mendelian randomization indicated JAM3 as having positive causal relationship (β = 0.1459, P = 0.0043) with schizophrenia. The well performance (AUC = 0.7930 (0.7908, 0.7951), G-mean = 0.6683 (0.6652, 0.6714)) of LightGBM model which consisted of protein and population features presented a strong and robust predictive ability of our findings. Furthermore, the investigation on biological functions of 31 proteins showed the enrichment in cell adhesion, integrin binding, chemotaxis and extracellular matrix organization through enrichment analyses.