Reshaping computational neuropsychiatry beyond synaptopathy.

Abstract

Computational neuropsychiatry is a leading discipline in explaining psychopathology in terms of neuronal message passing, distributed processing and belief propagation in neuronal networks. Active Inference (AI) is a way of representing this dysfunctional signal processing. According to the AI approach, all neuronal processing and action selection can be explained by maximizing Bayesian model evidence or minimizing variational free energy. Following these principles, it has been suggested that dysconnection in neuronal networks results in aberrant belief updating and erroneous inference, leading to psychiatric and neurologic symptoms. However, there is a classic distinction between disorders of inference (or synaptopathy-including the majority of psychiatric disorders) and disorders of brain function (including vascular neurological pathologies and severe forms of tauopathy and synucleinopathies). This distinction is generally based on the idea that synaptopathies impair neuromodulatory precision weighting, leading to rigid inferences or heightened sensitivity to noise, while disorders of brain function are linked to damage in the nervous system (disconnection). This makes it challenging to apply the logic of the free energy principle. We suggest that this distinction will enable future models of neuropsychiatric symptoms to be improved by considering more than neuronal message passing.