Immune activation has been recently found to play a large part in the development of schizophrenia, but the underlying mechanism remains largely unknown. Here, we report the construction of a high-quality protein interaction network for schizophrenia (SCZ Network) using a "neighborhood walking" approach to searching across human interactome network for disease-related neighborhoods. The spatiotemporal expression pattern of the immune genes in the SCZ Network demonstrates that this disease network is sensitive to the perturbation of immune activation during mid-to late fetal development and adolescence. The immune genes in the network are involved in pathways regulating the formation, structure and function of synapses and neural connections.Using single-cell transcriptome sequencing on the brains of immune-activated mice, we found that immune activation disturbed the SCZ network in the major brain cell types and the dysregulated pathways were also involved in synapse regulation, demonstrating that our "neighborhood walking" approach enables biological discovery in complex disorders. 1 in the selected neighborhoods for removing nodes from the candidate SCZ networks. Using on these optimal cutoffs, we obtained a SCZ Network, which has a seed ratio of 72.3% and contains 3,975 nodes and 92,318 edges, including 869 immune genes (658 known SCZ candidate genes) and 3,106 non-immune genes (2,215 known SCZ candidate genes) ( Fig. 1e, Supplementary Data 4).
The high quality SCZ network.The primary purpose of our "neighborhood walking" approach is to extract a high-quality disease network from human interactome by removing false positives of SCZ candidate genes while keeping the true positives and novel candidate genes in the network. Of the 3,437 SCZ candidate genes, 3,265 were mapped on the human interactome ( Fig. 2a). After the "neighborhood walking" process, 2,873 schizophrenia candidate genes and 1,102 novel candidate genes were kept in the SCZ Network (Fig. 1e). In order to verify that the "neighborhood walking" approach has removed false positives and kept true positives, we evaluated the SCZ relevance of the kept genes (genes inside the SCZ Network) and removed genes (genes outside the SCZ Network) using three datasets collected from literature (Supplementary Data 5): (i) dysregulated genes in the brain of schizophrenia patients 39-42 , (ii) variants associated with schizophrenia 43 and (iii) genes with brain-specific expression (Expression Atlas) 44 . Compared to removed genes, the kept genes showed significant enrichment with dysregulated genes in the brains of schizophrenia patients (p = 0.0007 for all kept genes, two-side Fisher's exact tests) ( Fig. 2b, Supplementary Data 5), SCZ-associated variants (p = 5.9827e-98 for all kept genes, p = 7.5082e-07 for kept seeds, p = 0.0183 for kept nonseeds, two-side Fisher's exact tests) ( Fig. 2c, Supplementary Data 5), and brain-specific genes (p = 2.3283e-99 for all kept genes, p = 6.0824e-11 for kept nonseeds, two-side Fisher's exact tests) ( Fig. 2d, Supplementary Dat...