Metabolomics strategy assisted by transcriptomics analysis to identify biomarkers associated with schizophrenia

Metabolomics strategy was perform to identify the novel serum biomarkers linked to schizophrenia with the assistance of transcriptomics analysis. Two analytical platforms, UPLC-Q-TOF MS/MS and 1H NMR, were used to acquire the serum fingerprinting profiles from a total of 112 participants (57 healthy controls and 55 schizophrenia patients). The differential metabolites were primarily selected after statistical analyses. Meanwhile, GSE17612 dataset downloaded from GEO database was implemented WGCNA analysis to discover crucial genes and corresponding biological processes. Based on metabolomics analysis, the metabolic distinctions were explored under the aid of transcriptomics. Then using Boruta algorithm identified the biomarkers, and LASSO regression analysis and Random Forest algorithm were used to evaluate the performance of the diagnostic model constructed by biomarkers selected. A total of four metabolites (α-CEHC, neuraminic acid, glyceraldehyde and asparagine) were selected as the biomarkers to establish diagnosis model. The performance of this model showed a higher accuracy rate to distinguish schizophrenia patients from healthy controls (area under the receive operating characteristic curve, 0.992; precision recall curve, 1.000, the mean accuracy of random forest algorithm, 95.00%). A four–biomarker model (α-CEHC, neuraminic acid, glyceraldehyde and asparagine) seems to be a good model for diagnosing schizophrenia patients. It might be helpful to guide the future studies on permitting early intervention designed to prevent disease progression.