Schizophrenia (SZ) and bipolar disorder (BD) are common and complex psychiatric disorders that have a lifetime prevalence of approximately 1%. Disorder-specific genetic factors as well as common genetic factors contribute to the pathogenesis of both disorders. To reduce the clinical and genetic heterogeneity between the two disorders and to differentiate between them, we considered hippocampal volume and cognitive function as promising intermediate phenotypes. Hippocampal volume and cognitive function are related to genetic factors and are more strongly impaired in SZ than BD. We are continuing research to develop a method to differentiate SZ and BD by polygenic risk score (PRS), hippocampal volume, and cognitive function using machine learning. To achieve this goal, we examined (i) subcortical volume changes between SZ and BD; (ii) causal relationships among SZ, BD, and intellectual impairments; and (iii) associations with genetic factors and intellectual function. We found that (i) hippocampal volumes were decreased in both patients with SZ and BD, whereas amygdala volumes were decreased only in patients with SZ; (ii) there was a bidirectional causal relationship between SZ and intellectual impairments, but not between BD and intellectual impairments; and (iii) genetic factors for differentiating SZ from BD were negatively correlated with intellectual function. These results suggest that amygdala volume is more useful for differentiating SZ from BD than hippocampal volume, and that cognitive impairments are also useful in terms of causal and genetic associations. We plan to develop a method to differentiate SZ from BD by combining PRS, amygdala volume, and cognitive function, and by applying machine learning.
 Authors' abstract