Abstract
Schizophrenia is a chronic illness that imposes a significant burden on patients, their families, and the health care system. While it has a substantial genetic component, its heterogeneous nature-both genetic and clinical-limits the ability to identify causal genes and mechanisms. In this study, we analyzed the blood transcriptomes of 398 samples (212 patients with schizophrenia and 186 controls) obtained from five public datasets. We demonstrated this heterogeneity by clustering patients with schizophrenia into two molecular subtypes using an unsupervised machine-learning algorithm. We found that the genes most influential in clustering were enriched in pathways related to the ribosome and ubiquitin-proteasomes system, which are known to be associated with schizophrenia. Based on the expression levels of these genes, we developed a logistic regression model capable of predicting schizophrenia samples in unrelated datasets with a positive predictive value of 64% (p value = 0.039). In the future, integrating blood transcriptomics with clinical characteristics may enable the definition of distinct molecular subtypes, leading to a better understanding of schizophrenia pathophysiology and aiding in the development of personalized drugs and treatment options.