Abstract
Schizophrenia (SCZ) is a complex psychiatric disorder presenting challenges for characterization. The current study aimed to identify and evaluate disease-responsive essential genes (DREGs) to enhance the molecular characterization of SCZ. RNA-sequencing data from PsychENCODE (536 SCZ patients, 832 controls) and peripheral blood transcriptome data from 144 recruited subjects (59 SCZ patients, 6 non-SCZ psychiatric patients, 79 controls) are analyzed. Shared differential expression genes are obtained using three algorithms. Support vector machine (SVM)-based recursive feature elimination is employed to identify DREGs. The biological relevance of these DREGs is examined through protein-protein interaction network, pathway enrichment, polygenic scoring, and brain tissue expression. Key DREGs are validated in SCZ animal models. A DREGs-based machine-learning model for SCZ characterization is developed and its performance is assessed using multiple datasets. The analysis identified 184 DREGs forming an interconnected network involved in synaptic plasticity, inflammation, neuronal development, and neurotransmission. DREGs exhibited distinct expression in SCZ-related brain regions and animal models. Their genetic contributions are comparable to genome-wide polygenic risk scores. The DREG-based SVM model demonstrated high performance (AUC 85% for SCZ characterization, 79% for specificity). These findings provide new insights into the molecular mechanisms underlying SCZ and emphasize the potential of DREGs in improving SCZ characterization.