人脑被视为自然界最复杂的系统，是认知功能的物质基础。研究表明绝大部分认知过程并非是由某个脑区单独活动完成的，而是多个脑区协同活动的结果。因此，观测和描绘脑区之间协同活动模式是理解认知过程的重要途径。功能磁共振成像作为一种无创在体神经活动观测技术，是观察和记录人脑活动状态的重要工具。结合复杂网络理论，研究者将脑区之间功能磁共振成像信号的统计依赖性定义为功能连接，构建了人脑功能连接组，为描绘和分析脑区之间协同活动模式开辟了新路径。研究发现人脑功能连接组包含拥有高度稠密功能连接的脑区，即核心脑区。核心脑区在维持人脑信息通讯和完成复杂认知任务中发挥重要作用，并且是多种神经精神疾病的优先攻击靶点。因此，准确识别核心脑区并描绘其功能连接图谱对理解人脑信息通讯、认知过程、神经精神疾病的发病机制等具有重要意义。然而，此前研究报告的核心脑区位置的一致性及可重复性很低。在群体水平一致存在且高度可重复的人脑功能连接组核心脑区图谱至今仍未建立。此外，尽管对于人脑功能连接组可遗传性的研究已经取得了重要的成果，但是核心脑区相较于非核心脑区是否有其独特的遗传特征却仍不清楚。阐明这些可能存在的遗传特征将有效地促进我们对于核心脑区如何在神经发育过程中产生、如何支撑复杂认知过程及如何成为神经精神疾病攻击靶点的进一步理解。为解决以上问题，本论文首先基于全球61个扫描中心的5212名健康青年被试的静息态功能磁共振成像数据的功能连接组协调荟萃分析识别了在群体水平一致存在且高度可重复的核心脑区。然后，本论文绘制了核心脑区的功能连接图谱并分析了核心脑区在维持全脑信息通讯中的独特作用。最后，结合艾伦研究所人脑图谱及人脑毕生发展图谱的转录组数据、机器学习及基因富集分析，本论文证明了人脑功能连接组核心脑区相较于非核心脑区具有其时空特异的转录模式。

识别群体水平一致存在的人脑功能连接组核心脑区(研究一)：本研究主要目的是识别在群体水平一致存在且高度可重复的人脑功能连接组核心脑区。本研究使用来自亚洲、欧洲、北美洲及澳洲的61个扫描中心的5212名健康青年被试(18至36岁，2377名男性)的静息态功能磁共振成像数据完成了首个全球范围的人脑功能连接组协调荟萃分析，在多个联合皮层脑区识别到核心脑区，其中最鲁棒的核心脑区位于外侧顶叶皮层。我们在这些核心脑区中识别到35个局部极大值。这些核心脑区富集于注意系统、默认模式系统及额顶系统。验证分析结果显示，这些核心脑区在群体水平是一致存在且高度可重复的。剔除单个扫描中心的数据对于核心脑区的空间分布无明显影响(Dice系数=0.990±0.006，范围：0.966-0.997；局部极大值核心脑区的位移绝大部分小于6毫米)，跨被试和扫描中心的联合分析结果均与协调荟萃分析识别到的核心脑区空间分布高度重合(Dice系数范围：0.867-0.924)。因此，本研究通过首个全球范围的人脑功能连接组协调荟萃分析识别了在群体水平一致存在且高度可重复的人脑功能连接组核心脑区。

人脑功能连接组核心脑区的功能连接图谱(研究二)：本研究主要目的是绘制人脑功能连接组核心脑区的功能连接图谱，并分析核心脑区在维持全脑信息通讯中的独特作用。首先，本研究使用研究一中的61个扫描中心的5212名健康青年被试(18至36岁，2377名男性)的静息态功能磁共振成像数据，通过体素水平的功能连接组协调荟萃分析绘制研究一识别的35个核心脑区的功能连接图谱，基于层级聚类分析将35个核心脑区分为三个类别。类别Ⅰ的21个核心脑区连接至注意系统、额顶系统及躯体感觉运动系统脑区；类别Ⅱ的4个核心脑区连接至视觉系统脑区及部分高级感觉通路脑区；类别Ⅲ的10个核心脑区连接至默认模式系统脑区及部分边缘系统脑区；类别Ⅰ及类别Ⅲ的核心脑区均连接至皮层下结构，但连接至不同的皮层下核团。结合核心脑区功能连接图谱的认知功能解码的结果，我们推测类别Ⅰ的核心脑区接收来自皮层下核团、联合视觉皮层及躯体感觉皮层的外部刺激信息，在注意系统、额顶系统及躯体感觉运动系统内及系统间进行信息整合及处理，完成决策及认知控制过程；类别Ⅱ的核心脑区将来自视觉系统的信息中转至更高层级的决策及认知控制相关的脑区；类别Ⅲ的核心脑区参与内部导向的认知过程。此外，类别Ⅲ的位于左脑外侧额中回后部8Av区域的核心脑区与默认模式系统及额顶系统脑区之间均有稠密的功能连接，表明核心脑区8Av是默认模式系统与额顶系统之间重要的连接子。结合非人灵长类8aV脑区连接模式的研究结果，我们进一步推测8Av脑区的默认模式系统与其它功能系统之间重要的连接子作用可能广泛存在于灵长类大脑中。因此，尽管核心脑区都有稠密的功能连接，但它们连接至不同的功能系统且都保有重要的系统间连接，从而在维持全脑信息通讯中承担独特的作用。

人脑功能连接组核心脑区的转录组特征(研究三)：本研究主要目的是探究人脑功能连接组核心脑区相较于非核心脑区是否有其独特的转录组特征。考虑到大量证据显示人脑功能连接组具有显著的可遗传性且受基因分型的影响，我们推测人脑功能连接组核心脑区相较于非核心脑区有其独特的转录组特征。我们使用基于机器学习的数据驱动分析方法探究这些潜在的转录组特征。首先，我们使用艾伦研究所人脑图谱的10027个基因的转录组数据训练机器学习分类器，有效地将1158个脑区分为核心脑区或非核心脑区(分类正确率65.3%-91.8%)。对分类结果有重要贡献的是150个与神经肽信号通路、神经发育过程、代谢过程及精神疾病相关的基因。然后，结合人脑毕生发展图谱的转录组数据，我们发现核心脑区相较于非核心脑区的神经发育过程及代谢过程相关基因的转录水平的发育轨迹有明显差异。最后，我们使用已发表的人脑多模态神经影像数据对以上转录组数据分析结果的验证分析得到了一致结论。因此，人脑功能连接组核心脑区相较于非核心脑区具有其时空特异的转录模式，涉及的基因富集于神经肽信号通路、神经发育过程、代谢过程及精神疾病。结合已有研究结果，我们推测这一时空特异的转录模式可能与核心脑区的发育产生、支撑复杂认知功能及在神经精神疾病中的脆弱性密切相关。

The human brain is arguably the most complex system in biology, serving as the basis of our cognitive function. Studies demonstrated that virtually all domains of cognitive function require coordinative activity of distributed brain regions rather than independent activity. Thus, quantitatively measuring and describing the coordinative activity of brain regions is essential for understanding cognitive process. Functional magnetic resonance imaging (fMRI), a noninvasive in vivo neural activity measuring technique, is combined with complex network theory to construct the functional connectome by defining the statistical dependence between fMRI signals of brain regions as functional connectivity, which provides new route to quantitatively measure and describe the coordinative activity pattern of brain regions. Macroscopic functional connectomic analyses have identified sets of densely connected regions in the human brain, known as connectome hubs, which play a vital role in global brain communication, cognitive processing, and are targeted attacked by neuropsychiatric disorders. Therefore, accurately localizing these hubs and mapping their functional connectivity profiles are indispensable for understanding global brain communication, cognitive processing, and pathophysiological mechanisms underlying neuropsychiatric disorders. However, anatomical locations of these hubs are largely inconsistent and less reproducible among extant reports. The consistency and reproducibility of functional connectome hubs’ anatomical localization have not been established to date. Moreover, the genetic signatures underlying robust hubs remain unknown. Elucidating these genetic signatures will benefit our understanding of how functional connectome hubs emerge in neurodevelopment, function in complex cognition, and are involved in neuropsychiatric disease. To address these issues, this thesis conducted the first worldwide, harmonized meta-connectomic analysis by pooling resting-state fMRI (rsfMRI) data across 5,212 subjects from 61 international cohorts, identified highly consistent and reproducible functional connectome hubs, mapped their functional connectivity profiles, and analyzed their distinctive roles in bridging brain networks. Using post-mortem transcriptome data from the Allen Human Brain Atlas and the BrainSpan Atlas，machine learning, and Gene Ontology enrichment analysis, we demonstrated that these connectome hubs have a spatiotemporally distinctive transcriptomic pattern in contrast to non-hubs.
Identifying Consistent Functional Connectome Hubs of the Human Brain (Study 1): The aim of this study is to identify highly consistent and reproducible functional connectome hubs. We provided the first worldwide harmonized meta-connectomic analysis of functional connectome hubs by pooling a large-sample rsfMRI dataset of 5,212 healthy young adults (aged 18-36 years, 2,377 males) across 61 cohorts from Asia, Europe, North America, and Australia. We identified functional connectome hubs in multiple association regions, with the most robust findings occurring in several lateral parietal regions, and indentified 35 local peaks in these connectome hubs. These functional connectome hubs were enriched in the attention, default-mode, and frontoparietal networks. Validation analysis demonstrated that these functional connectome hubs are highly consistent and reproducible both across subjects and across cohorts. Specifically, leave-one-cohort-out had no apparent effect on the localization of these hubs indicated by a high Dice’s coefficient (mean±sd: 0.990±0.006; range: 0.966-0.997) and very few displacements of hub peaks (mostly fewer than 6 mm); functional connectome hubs obtained by harmonized meta-connectomic analysis were largely overlapped with those obtained by conjunction analysis across subjects and cohorts (Dice：0.867-0.924). Together, through the first worldwide harmonized meta-connectomic analysis of 5,212 subjects across 61 cohorts we identified highly consistent and reproducible functional connectome hubs.
Mapping Functional Connectivity Profiles of Brain Hubs (Study 2): The aim of this study is to map robust functional connectivity profiles of the 35 hubs identified in Study 1 and to further analyze their distinctive roles in bridging brain networks. Using preprocessed rsfMRI data of 5,212 healthy young adults (aged 18-36 years, 2,377 males) from 61 cohorts in Study 1, we conducted a seed-to-whole-brain connectivity meta-analysis in a harmonized protocol again and obtained robust voxel-wise functional connectivity profiles for each hub region. Hierarchical clustering analysis on these connectivity profiles clearly divided the 35 hub regions into three clusters. Cluster I consists of 21 hubs that are connected with extensive areas in the dorsal attention, ventral attention, frontoparietal, and somatomotor networks. Cluster Ⅱ consists of four hubs that are densely connected with the visual network and brain regions involved in the sensorimotor pathway. Cluster Ⅲ consists of 10 hubs that have robust connections with the default-mode network and limbic network. Both Cluster I and Ⅲ hubs are connected with subcortical structure, but they are connected with different subcortical nuclei. Combined with cognitive function decoding results of these hubs’ functional connectivity profiles, we speculated that Cluster I hubs receive the information of external stimulus from subcortical structures and sensory and visual cortices, process and integrate this information within and between the attention, frontoparietal, and somatomotor networks, and complete the decision making and cognitive control functions; Cluster Ⅱ hubs transfer information from the visual system to higher-level cortices related to decision making and cognitive control; Cluster III hubs engage in internally directed cognition. Of particular interest is that within Cluster III, a left posterior middle frontal hub called ventral area 8A (8Av) shows robust functional connections with both the default-mode and frontoparietal regions, suggesting its connecter role between the default-mode and frontoparietal networks. Combined with previous findings of 8aV region’s connectivity pattern in the non-human primate brain, we futher speculated that 8Av region’s connecter role between the default-mode and other networks may widely exist in the primate brain. Collectively, whereas all hubs possess dense connections, they are connected with different brain networks and retain significant internetwork connections to play distinctive roles in bridging communications among brain networks, which preserves efficient communication across the whole brain feasible.
Transcriptomic Signatures Associated with the Robust Connectome Hubs (Study 3): The aim of this study is to verify whether the robust functional connectome hubs identified in Study 1 have distinctive transcriptomic signatures intrast to non-hubs. Considering abundant evidence of significant heritability of human brain functional connectivity and substantial evidence supporting genotypic variation involved in functional connectome architectures of the human brain, we reasoned that functional connectome hubs have distinctive transcriptomic signatures intrast to non-hubs. We conducted a data-driven analysis using machine learning approaches to explore these potential transcriptomic signatures. First, machine-learing classifiers trained using transcriptomic features of 10,027 genes from the Allen Human Brain Atlas effectively classified 1,158 brain samples as hubs or non-hubs (classification accuracy: 65.3%-91.8%), which was mainly contributed by 150 genes involved in the neuropeptide signaling pathway, neurodevelopmental processes, metabolic processes, and psychiatric disorders. Then, using transcriptomic data form the BrainSpan Atlas, we observed pronounced diverging transcriptomic trajectories between hub and non-hub regions for genes associated with neurodevelopmental processes and metabolic processes. Finally, validation analysis using multimodal neuroimaging data from other independent laboratories conformed these transcriptomic analysis results. Taken together, intrast to non-hubs, the robust functional connectome hubs have a spatiotemporally distinctive transcriptomic pattern dominated by genes involved in the neuropeptide signaling pathway, neurodevelopmental processes, metabolic processes, and psychiatric disorders. Combined with previous findings, we speculated that this spatiotemporally distinctive transcriptomic pattern is relevant to functional connectome hubs’ neurodevelopment, supporting complex cognitive functions, and susceptibility to neuropsychiatric disorders.