Computational ontology of brain systems across the human neuroimaging literature

项目摘要/摘要 精神疾病的基于症状的诊断是高度并存的，在生物学上是不同的，而且很差 对治疗反应的预测。国家精神卫生研究所领导了重新定义精神卫生的努力 疾病的生物学原因，建立研究领域标准(RDoC)框架作为指导 研究基本大脑系统的变异。RDoC一直很有影响力，在数百项赠款和 出版物，但尚待系统验证。目前尚不清楚电路功能是否与潜在的 RDoC的大脑系统在研究中是可重现的，组织原则在很大程度上仍未经过测试。 虽然RDoC作为模块化层级的结构在静态分析中有证据，但它并没有 显示了这是否适用于支持受精神疾病影响的不同精神状态的系统。 有必要对RDoC进行验证，并为以下项目建立基本的组织原则 由人脑结构和功能共同定义的系统。这项建议的目标是将大量的- 扩展计算神经成像元分析以构建数据驱动的本体，该本体不仅将用作 评估RDoC有效性的基准，但也表征了人脑系统的体系结构 功能。长期目标是通过大脑内健康功能的不同来重新定义精神疾病。 数据驱动的本体论系统，促进神经调节治疗的合理靶向。 拟议的荟萃分析将是该领域中最全面的，拥有18,155项核磁共振和正电子发射计算机断层扫描研究 已经收好了。这些研究中考虑的心理功能是从文章文本中提取出来的 使用自然语言处理，大脑回路将从大脑坐标数据中映射出来 据报道。假设大脑系统是由可复制的电路-功能链接组成的 进入模块化层次结构，这对于某些系统将需要更新RDoC。这将通过以下方式进行测试 将RDoC系统与数据驱动本体的系统进行比较。目标1：电路功能的重复性 链接将通过神经网络分类器预测功能在文章文本中的性能进行评估 脑扫描数据中的电路，反之亦然。目标2：大脑系统的模块化将通过图表进行评估 理论方法，层级结构将通过代表性相似性分析进行评估。 该项目的影响将是验证最重要的精神病学研究框架并确定 人类大脑系统通过一种创新的计算策略。与有针对性的学术培训一起 在神经生物学方面，该奖学金旨在为成为一名领先的内科科学家提供职业准备。 计算精神病学的进展。培训将得到世界级环境的支持 与精神病学、神经科学和计算机科学方面受人尊敬和积极参与的导师一起提供计算资源。

Project Summary/Abstract Symptom-based diagnoses of mental illness are highly comorbid, biologically heterogeneous, and poorly predictive of treatment response. The National Institute of Mental Health has led efforts to redefine mental illness by its biological causes, establishing the Research Domain Criteria (RDoC) framework as a guide for investigating variation in basic brain systems. RDoC has been influential, named in hundreds of grants and publications, but it has yet to be systematically validated. It is unknown whether circuit-function links underlying the RDoC brain systems are reproducible across studies, and organizing principles remain largely untested. While the structure of RDoC as a modular hierarchy has evidence in resting state analyses, it has not been shown whether this applies to systems that support the diverse mental states affected in psychiatric disease. It is necessary to validate RDoC, and moreover, to establish fundamental principles of organization for systems defined jointly by human brain structure and function. The objective of this proposal is to apply large- scale computational neuroimaging meta-analyses to build a data-driven ontology that will not only serve as a benchmark in evaluating the validity of RDoC but also characterize the architecture of systems for human brain function. The long-term goal is to redefine mental illness by differences from healthy function within the brain systems of a data-driven ontology, facilitating rational targeting of neuromodulation treatments. The proposed meta-analyses will be the most comprehensive in the field with 18,155 MRI and PET studies already collected. The mental functions considered in these studies have been extracted from article texts using natural language processing, and brain circuits will be mapped from the brain coordinate data that were reported. The hypothesis is that brain systems are comprised of reproducible circuit-function links organized into a modular hierarchy, which for some systems will require updates to RDoC. This will be tested by comparing RDoC systems against those of a data-driven ontology. Aim 1: The reproducibility of circuit-function links will be evaluated by the performance of neural network classifiers predicting functions in article texts from circuits in brain scan data, and vice versa. Aim 2: The modularity of brain systems will be evaluated by a graph theoretic approach, and hierarchical structure will be assessed by representational similarity analysis. The impact of this project will be to validate the foremost psychiatry research framework and to characterize human brain systems through an innovative computational strategy. Together with targeted academic training in neurobiology, the fellowship is designed to offer preparation for a career as a physician-scientist leading advances in computational psychiatry. Training will be supported by an environment that combines world-class computing resources with esteemed and engaged mentors in psychiatry, neuroscience, and computer science.