Mixture Model Clustering of Phenotype Features Reveals Evidence for Association of DTNBP1 to a Specific Subtype of Schizophrenia

Background: While DTNBP1, DISCI, and NRG1 have been extensively studied as candidate genes of schizophrenia, results remain inconclusive. Possible explanations for this are that the genes might be relevant only to certain subtypes of the disease and/or only in certain populations. Methods: We performed unsupervised clustering of individuals from Finnish schizophrenia families, based on extensive clinical and neuropsychological data, including Structured Clinical Interview for DSM-IV (SCID) information. Families with at least one affected member with DSM-IV diagnosis of a schizophrenia spectrum psychosis were included in a register-based ascertainment. Final sample consisted of 904 individuals from 288 families. We then used the cluster phenotypes in a genetic association study of candidate genes. Results: A robust three-class clustering of individuals emerged: 1) psychotic disorder with mood symptoms (n = 172),2) core schizophrenia (n = 223), and 3) absence of psychotic disorder (n = 509). One third of the individuals diagnosed with schizophrenia were assigned to cluster 1. These individuals had fewer negative and positive psychotic symptoms and cognitive deficits but more depressive symptoms than individuals in cluster 2. There was a significant association of cluster 2 cases with the DTNBP1 gene, while the DISCI gene indicated a significant association with schizophrenia spectrum disorders based on the DSM-IV criteria. Conclusions: In the Finnish population, DTNBP1 gene is associated with a schizophrenia phenotype characterized by prominent negative symptoms, generalized cognitive impairment, and few mood symptoms. Identification of genes and pathways related to schizophrenia necessitates novel definitions of disease phenotypes associated more directly with underlying biology.