Integrative analysis of multiomic datasets for discovery of molecular underpinnings of large-scale human brain networks

摘要 脑成像计划正在获得越来越大和全面的神经成像和多组学- 例如基因组数据集和转录组数据集。人类神经科学中对这类数据的现有分析倾向于 一方面，搜索认知、行为或疾病与基因组、转录- 另一种是大部头或大脑形态和连通性。这些有价值的分析稳步推进了我们的 关于人脑功能的知识。但它们也在我们对这一功能如何发挥作用的理解上留下了一个关键的空白- 大脑进化、发育和组织的相互作用产生了动因。 本提案将通过整合几个大而不同获得的神经成像来帮助填补这一空白 和多组数据集。它将通过将日益丰富的数据可用性与现代统计相结合来实现这一点 方法，并与其研究人员在网络神经科学方面的互补专业知识，计算生物学- OGY、人类进化、数据协调、认知发展和老化神经科学。 该提案将把可遗传表达与大脑几个关键区域的大脑网络表型联系起来 数以千计的基因和成千上万的个体。它将通过采用和应用可遗传的模型来做到这一点 基因表达(根据基因组织表达项目获得的转录数据进行培训，并与之相关 项目)到人类连接组项目和英国生物库获得的神经成像基因组数据。 然后，该提案将区分适应性和非适应性脑网络表型。它会这么做的 通过量化这些表型在最近人类进化中的自然选择，使用来自 古人类，以测试与大脑网络变异相关的基因的自然选择压力。 该提案将最终描述可遗传基因表达和网络表型之间的关系 在典型和非典型的发育和衰老中。它将通过将可遗传的基因表达从大的 由青少年大脑认知发展研究获得的神经成像基因组数据集，CaM- 老年和神经科学桥梁中心和阿尔茨海默病神经成像倡议。它将链接到 这种表达的变化与发育和衰老过程中脑网络表型的变化有关，并将 描绘精神病谱系症状或认知障碍的基因表达脑网络特征。 综上所述，该提案整合了大规模脑网络的演进、发展和组织- 行得通。具体地说，它首次将基因表达和脑网络表型联系在一起，跨越了几个不同的领域。 在许多个体中发现了神经元，并以这种方式为神经成像基因组学开辟了一个新的方向。该建议书广告- 万斯通过分析加强现有的基因组、转录和神经成像来发现神经科学-- 老化数据。最后，通过传播作为这些分析的一部分创建的所有软件和结果，PRO- POSAL最终加速了未来大型神经科学数据集的严格和可重复的整合。

SUMMARY Brain-mapping initiatives are acquiring increasingly large and comprehensive neuroimaging and multiomic— e.g. genomic and transcriptomic—datasets. Existing analyses of such data in human neuroscience tend to search for links between cognition, behavior or disease on the one hand, and properties of genomes, transcrip- tomes or brain morphology and connectivity on the other. Such valuable analyses have steadily advanced our knowledge of human brain function. But they have also left a critical gap in our understanding of how this func- tion arises from the interplay of brain evolution, development and organization. The present proposal will help fill this gap by integrating several large and disparately acquired neuroimaging and multiomic datasets. It will do so by combining the increasing availability of rich data, with modern statistical methods, and with complementary expertise of its investigators in network neuroscience, computational biol- ogy, human evolution, data harmonization, and cognitive developmental and aging neuroscience. The proposal will link heritable expression to brain-network phenotypes across several key brain regions for thousands of genes and in thousands of individuals. It will do so by adopting and applying models of heritable gene expression (trained on transcriptomic data acquired by the Gene Tissue Expression Project, and allied projects) to neuroimaging genomic data acquired by the Human Connectome Project and the UK Biobank. The proposal will then distinguish between adaptive and non-adaptive brain-network phenotypes. It will do so by quantifying the natural selection of these phenotypes in recent human evolution, using ancient DNA from archaic hominins, to test for natural-selection pressures on genes associated with brain-network variation. The proposal will finally delineate the relationship between heritable gene expression and network phenotypes in typical and atypical development and aging. It will do so by imputing heritable gene expression from large neuroimaging genomic datasets acquired by the Adolescent Brain Cognitive Development Study, the Cam- bridge Centre for Ageing and Neuroscience, and the Alzheimer's Disease Neuroimaging Initiative. It will link the variation in this expression to the variation of brain-network phenotypes in development and aging, and will delineate gene-expression brain-network signatures of psychosis-spectrum symptoms or cognitive impairment. Collectively, the proposal integrates the evolution, development, and organization of large-scale brain net- works. Specifically it links, for the first time, gene expression and brain-network phenotypes across several re- gions in many individuals, and in this way opens a new direction in neuroimaging genomics. The proposal ad- vances discovery neuroscience through analyses that enhance existing genomic, transcriptomic and neuroim- aging data. Finally, through dissemination of all software and results created as part of these analyses, the pro- posal ultimately accelerates future rigorous and reproducible integration of large neuroscience datasets.