Examining the hierarchical structure of the RDoC framework using large-scale data-driven computational approaches

点击翻译按钮获取中文摘要

Project Summary The Research Domain Criteria (RDoC) applies an integrative, dimensional approach anchored in circuit neuroscience, genes, molecules, and behaviors. The RDoC framework, currently only for research, ultimately aims at facilitating the development of psychiatric nosology (disorder-classification system) based upon primary behavioral functions and their associated biological features that the brain has evolved to carry out. Although the impetus behind RDoC is in the right direction, for greater efficacy of RDoC in clinical translation, a data-driven examination is needed to validate and refine the architecture of RDoC. Further, several key questions remain unanswered. First, as noted in the current RFA (RFA-MH-19-242), since the inception of RDoC, a thorough data-driven validation that broadly explores, compares, and validates the constructs within the framework has not been performed. Second, to increase clinical translation of the RDoC framework, it is essential to assess whether constructs within a domain consistently relate to similar dimensions of psychopathology. Thus, providing data-driven evidence for the convergent and discriminant validity of the RDoC framework in predicting psychopathology. Lastly, and perhaps more fundamentally, it is unclear whether carefully crafted behavioral paradigms are required to examine domain-specific features (behavioral or circuit- level) or task-free paradigms (e.g., resting-state) can be computationally employed to extract similar domain- specific features. The lack of task instructions in resting-state paradigms enhances compliance in clinical populations, makes data aggregation across sites straightforward, and could provide a higher cost-benefit ratio if a single resting-state scan can provide information that would otherwise require multiple, carefully crafted, domain-specific neuroimaging task scans. Here, we propose to mine, systemically and computationally, three large-scale datasets from the general population and diagnosed patient populations to answer critical questions regarding the validity of the RDoC framework. Specifically, we aim to examine whether: (1) within- domain constructs overlap more than do between-domain constructs; (2) within-domain constructs relate to similar dimensions of psychopathology; and (3) task-free paradigms (e.g., resting-state) can be mined to extract similar domain-specific information that is usually extracted using specific task-based paradigms. By addressing these three key questions, our central goal is to provide the much-needed bottom-up examination of the RDoC framework to pave a pathway for its refinement and translation. Our long-term goal is to develop new computational frameworks to generate converging insights for grounding psychiatric nosology in biological features. Altogether, without careful data-driven validation, the RDoC framework remains theoretical. Hence, we advocate for developing a computational backbone for the RDoC framework to validate the assumptions underlying RDoC and facilitate framework refinement for greater clinical translation.