Neurocomputational Modeling Core

点击翻译按钮获取中文摘要

PROJECT SUMMARY The overarching goal of the Neurocomputational Modeling Core is to provide a common formal framework that can incorporate measures of neural activity, connectivity, and behavior across Projects 1-5 to (a) quantify the functional roles of the OCDnet and its components in approach-avoidance decision-making and OCD symptomatology, and (b) predict changes in decision-making dynamics and symptom severity as a result of neural and clinical interventions. To achieve these goals, we will leverage (a) models of decision dynamics and their modulation by neural activity within individual circuit nodes, and (b) graph-theoretic models of interactions across circuit nodes. To quantify decision dynamics during the PAAT task, we will use hierarchical Bayesian parameter estimation of the drift diffusion model (HDDM), which enables reliable estimation of decision parameters and their modulation by trial-by-trial variance in neural signals, and supports Bayesian hypothesis testing for how these parameters may differ as a function of clinical status and neuromodulation. We have previously shown how such “computational biomarkers” can discriminate between patient conditions and symptoms better than traditional measures of behavior and brain activity, including in an approach-avoid context. We will test how PAAT choices are modulated by a combination of task variables (e.g., rewarding and aversive outcomes), neural activity across OCDnet nodes, and OCD symptom severity. Preliminary results show that the HDDM captures expected differences in choice dynamics (e.g., choice bias) between patients and healthy controls. To quantify task-related functional interactions across this circuit, we will use ancestral graph models, which measure the strength and direction of information flow across graph nodes. We will use this combination of modeling approaches to test for changes in decision and circuit dynamics resulting from targeted interventions (e.g., lesions, stimulation, treatment). Machine learning methods will quantify the degree to which such quantitative model fitting improves (1) classification of patient condition and (2) our ability to map changes in behavior, circuit dynamics, and disease course following interventions. Building on our extensive experience in neural networks and levels of computation involved in motivated learning and decision making across species, our computational framework will facilitate not only enhanced sensitivity to discriminate between clinical conditions, but will also identify hypotheses about the likely mechanisms involved, which will be tested via causal manipulations using the same quantitative framework. Contribution to Overall Center Goals & Interactions with Other Center Components. Our modeling framework will be applied to data across all Projects, including measures of connectivity (P1), behavioral and neural activity (P2-5), clinical measures (P3- 5), and influences of neural (P2&5) and behavioral (P4) interventions. Cores B & C will help with localization and estimation of neural activity. We will benefit from interactions amongst experts with complementary expertise in systems and cognitive neuroscience, psychiatry, engineering, and computational modeling.