Discovering prognostic neuroimaging biomarkers of the psychosis spectrum using network control theory

点击翻译按钮获取中文摘要

PROJECT SUMMARY Schizophrenia, and other disorders of the psychosis spectrum (PS), most commonly emerge throughout development and are thought to be caused by disruptions to normative brain maturation occurring during this time. Critically, deviations from normative neurodevelopment are thought to precede the emergence of clinically significant PS symptoms by several years, highlighting the profound impact that their discovery would have for psychiatry research; if we can successfully identify the antecedent brain abnormalities, then we may be able to intervene early and reduce the risk of individuals developing schizophrenia. Uncovering these antecedent brain abnormalities requires predictive models built upon recent advances in network neuroscience and machine learning; moreover, such models must be coupled with large samples of longitudinal neuroimaging and clinical data to uncover truly prognostic biomarkers. Finally, sex differences are found in both the PS symptoms and neurodevelopment. Thus, studies that provide a precise understanding of how sex interacts with PS symptoms and abnormal neurodevelopment are needed. The purpose of the current study is to use advanced tools from network neuroscience and machine learning coupled with multi-modal neuroimaging to uncover biomarkers that can predict the emergence of PS symptoms throughout development. To achieve this goal, we will draw on multiple largescale cross-sectional and longitudinal neurodevelopmental datasets, including the Philadelphia Neurodevelopmental Cohort, the Healthy Brain Network, and the Adolescent Brain Cognitive Development study, to study brain structure and connectivity. We use Network Control Theory (NCT) to study connectivity. NCT treats the brain as a dynamical system allowing us to probe a region's capacity to control changes in brain states via white matter pathways. Compared to graph theory, NCT is a contemporary approach that posits an explicit model of how the brain's structure informs and constrains its function, enabling mechanistic insight into the dysconnectivity associated with the PS. We will quantify developmental abnormalities in NCT metrics using a nascent machine learning technique known as normative modeling. A normative model builds a growth chart of brain development that incorporates the expected variation in the relationship between age and the brain into its predictions. Then, deviations from these growth charts can be understood in terms of what is and what is not expected in a normative population. Here, we will build cross-sectional (Aim 1) and longitudinal (Aim 2) normative models of NCT metrics and use multivariate deviations to predict PS symptoms out-of-sample. Finally, Aim 3 will investigate the extent to which deviations from normative neurodevelopment mediate the relationship between sex and PS symptoms. The goal of this Pathway to Independence award is to build on my strong background in psychiatry, multimodal neuroimaging, network neuroscience, and machine learning by expanding my expertise to developmental psychopathology, NCT, and longitudinal neuroimaging data.