PROJECT SUMMARY There is an acute need in neuropsychiatric research to characterize the dimensional heterogeneity across patients within a categorical disorder such as schizophrenia (SCZ). It is unknown how clinically relevant individual differences in SCZ, such as the severity of cognitive deficits, relate to differences in underlying neural disturbances. Cognitive impairments in SCZ are hypothesized to involve widespread “dysconnectivity,” i.e., abnormal communication or interactions among brain regions in large-scale cortical networks. Noninvasive neuroimaging has revolutionized our understanding of systems-level connectivity disturbances in SCZ, yet the underlying cellular-level mechanisms remain unclear. A leading hypothesis for neuropathology in SCZ proposes disruptions in the balance between excitation (E) and inhibition (I) in cortical circuitry. This mechanism is supported by pharmacological models of SCZ which are hypothesized to induce synaptic disinhibition in cortex through antagonism of NMDA receptors. An emerging approach to bridge this explanatory gap—between neuroimaging observations and underlying biological processes—is to harness computational models of large-scale brain circuits that incorporate key features of neuronal and synaptic dynamics, thereby allowing mechanistic examination of how cellular-level disruptions propagate upward to produce systems-level dysfunction. The overarching goal of this “Computational Psychiatry” proposal is to develop a biophysically-based modeling framework that captures large-scale cortical dynamics at the individual-subject level, apply it to characterize dysconnectivity in SCZ and pharmacological manipulation, and relate model parameters to cognitive function. In Aim 1, we will develop and validate a model fitting framework that optimizes synaptic parameters in the model to match a subject’s personalized resting-state functional connectivity pattern, constrained by their own structural connectivity. Model parameters govern neurobiologically important synaptic properties such as E/I balance. To develop and apply this framework, we will leverage two existing, state-of-the-art multimodal neuroimaging datasets. The first dataset, from the Human Connectome Project (HCP), is from a large number of subjects, and will be used to characterize neural variation in the healthy population. The second dataset, collected at Yale and harmonized with HCP pipelines, is from patients with SCZ, and from matched healthy controls administered a subanesthetic dose of the NMDA receptor antagonist ketamine. In Aim 2, we will extend the model in two targeted directions that are grounded in known neurobiology and related to cortical dysfunction in SCZ: heterogeneity in local recurrent strength across the cortical hierarchy; and network-specific long-range interactions, which may be net-inhibitory. These extensions will be fit quantitatively at the individual level in both datasets. In Aim 3, we will relate model parameters to performance in working memory, a neurocognitive function central to cognitive deficits in SCZ, as well as to other non-imaging measures. This Computational Psychiatry research program advances our understanding of cortical disturbances in SCZ, illuminates individual variation in health and in SCZ, and establishes an extensible framework for Computational Psychiatry to link synapse-level hypotheses with human neuroimaging. Moreover, by combining clinical neuroimaging, pharmacology, and computational neuroscience, this study proposes a framework that can inform rational development of novel, personalized treatments designed to target specific neural disturbances.

PROJECT NARRATIVE Mechanisms underlying neural dysconnectivity in schizophrenia, and heterogeneity of the disorder across patients, are not well understood, due in part to explanatory gaps between noninvasive neuroimaging effects and synaptic-level hypotheses. This proposal aims to bridge these gaps through development of a computational modeling framework that uses biophysically- based models of large-scale cortical circuits to quantitatively fit brain structure and function at the individual-subject level — leveraging existing multimodal neuroimaging datasets from a large sample of healthy controls, from healthy controls administered a pharmacological model of schizophrenia, and from patients with schizophrenia. In doing so, this project has the potential to inform putative synaptic mechanisms that underlie dysconnectivity in schizophrenia, ultimately guiding development of novel therapies that are personalized for individual patients.