Developing in silico/in vitro models of schizophrenia positive symptoms

摘要

Schizophrenia is a mental illness characterized by the positive symptoms of hallucinations and delusions. In the field of computational psychiatry, a recently proposed hypothesis is that schizophrenia is caused by pathological changes in the recognition model due to elevated dopamine (DA) levels. Here, we constructed recognition models within a mathematical model (in silico) and a cultured nervous system (in vitro) in an attempt to validate the recognition model hypothesis and elucidate the mechanisms of positive symptoms in schizophrenia. For the mathematical model, analytical and numerical calculations showed that it is possible to extract the first principal component of inputs using spike-timing dependent plasticity and that its accuracy decreases as DA levels increase. Moreover, we demonstrated that this attenuation occurred even for extraction of natural images. Furthermore, for the in vitro model, we observed that cultured neural networks could not adequately separate hidden sources in the presence of DA, while they could do it in the absence. These results imply that the in silico and in vitro models can explain the mechanism of attenuation of sensory perception, such as hallucinations, at the neural network level and can be used as models for schizophrenia treatment.