Identification of Immune-Related Biomarkers of Schizophrenia in the Central Nervous System Using Bioinformatic Methods and Machine Learning Algorithms.

Abstract

Schizophrenia is a disastrous mental disorder. Identification of diagnostic biomarkers and therapeutic targets is of significant importance. In this study, five datasets of schizophrenia post-mortem prefrontal cortex samples were downloaded from the GEO database and then merged and de-batched for the analyses of differentially expressed genes (DEGs) and weighted gene co-expression network analysis (WGCNA). The WGCNA analysis showed the six schizophrenia-related modules containing 12,888 genes. The functional enrichment analyses indicated that the DEGs were highly involved in immune-related processes and functions. The immune cell infiltration analysis with the CIBERSORT algorithm revealed 12 types of immune cells that were significantly different between schizophrenia subjects and controls. Additionally, by intersecting DEGs, WGCNA module genes, and an immune gene set obtained from online databases, 151 schizophrenia-associated immune-related genes were obtained. Moreover, machine learning algorithms including LASSO and Random Forest were employed to further screen out 17 signature genes, including GRIN1, P2RX7, CYBB, PTPN4, UBR4, LTF, THBS1, PLXNB3, PLXNB1, PI15, RNF213, CXCL11, IL7, ARHGAP10, TTR, TYROBP, and EIF4A2. Then, SVM-RFE was added, and together with LASSO and Random Forest, a hub gene (EIF4A2) out of the 17 signature genes was revealed. Lastly, in a schizophrenia rat model, the EIF4A2 expression levels were reduced in the model rat brains in a brain-regional dependent manner, but can be reversed by risperidone. In conclusion, by using various bioinformatic and biological methods, this study found 17 immune-related signature genes and a hub gene of schizophrenia that might be potential diagnostic biomarkers and therapeutic targets of schizophrenia.