Data-driven validation of cognitive RDoC dimensions using deep phenotyping

点击翻译按钮获取中文摘要

Project Summary The NIMH research domain criteria (RDoC) reconceptualizes mental health research along a series of key cross-disorder dimensional constructs. However, these dimensions were determined in a top-down fashion by relatively small groups of researchers. We propose a data-driven approach that tests the validity of the key RDoC constructs of attention, cognitive control, and working memory. We will evaluate these constructs using multiple cognitive tasks per construct to examine their relationship to brain networks and their ability to predict real-world behaviors that are relevant to mental health. Finally, we propose an augmentation to the RDoC framework by adding new units of analysis: contrasts and practice. The current RDoC matrix maps directly from task paradigms to constructs and subconstructs, which is problematic because supposedly distinct constructs can sometimes map to exactly the same set of tasks. To address this, we propose a new RDoC unit of analysis called a “contrast”, which better reflects the usual logic of experimental design. We will identify mappings between cognitive systems constructs and contrasts through consultation with domain experts. We will then acquire a large-scale dataset to test both exploratory and confirmatory models for RDoC cognitive system constructs. Finally, we will evaluate whether these RDoC cognitive systems constructs are predictive of related real-world outcomes. The RDoC matrix links constructs to both behavioral measures and neural circuits, but the present mappings between cognitive systems constructs and brain systems are sparse and inconsistent. We will use a dense- sampling fMRI acquisition of 65 subjects each completing 10 scanning sessions on the same battery of tasks as the behavioral study, to develop a precise data-driven atlas of neural engagement at each level of the matrix, from contrasts to subconstructs to constructs. We will then validate the behaviorally-derived models using neural data, both between subjects and within subjects. We will also perform fully exploratory analyses to identify whether the data-driven neural circuit structure on these tasks diverges from the RDoC matrix. A long history of research in both and animals has shown that repeated practice on a task changes the way that the task is performed and the brain systems that support performance. We will leverage our behavioral and brain imaging samples to evaluate whether the structure of the cognitive systems domain remains constant with practice. In parallel we will also apply exploratory methods to assess the consistency of structural models estimated either early in training or after extensive practice. Overall, this project expands the RDoC matrix with two new units of analysis (contrasts and practice), and validates the constructs of attention, cognitive control, and working memory across both behavior and neural circuits.