Table_1_From Matrices to Knowledge: Using Semantic Networks to Annotate the Connectome.XLS

The connectome is regarded as the key to brain function in health and disease. Structural and functional neuroimaging enables us to measure brain connectivity in the living human brain. The field of connectomics describes the connectome as a mathematical graph with its connection strengths being represented by connectivity matrices. Graph theory algorithms are used to assess the integrity of the graph as a whole and to reveal brain network biomarkers for brain diseases; however, the faulty wiring of single connections or subnetworks as the structural correlate for neurological or mental diseases remains elusive. We describe a novel approach to represent the knowledge of human brain connectivity by a semantic network – a formalism frequently used in knowledge management to describe the semantic relations between objects. In our novel approach, objects are brain areas and connectivity is modeled as semantic relations among them. The semantic network turns the graph of the connectome into an explicit knowledge base about which brain areas are interconnected. Moreover, this approach can semantically enrich the measured connectivity of an individual subject by the semantic context from ontologies, brain atlases and molecular biological databases. Integrating all measurements and facts into one unified feature space enables cross-modal comparisons and analyses. We used a query mechanism for semantic networks to extract functional, structural and transcriptome networks. We found that in general higher structural and functional connectivity go along with a lower differential gene expression among connected brain areas; however, subcortical motor areas and limbic structures turned out to have a localized high differential gene expression while being strongly connected. In an additional explorative use case, we could show a localized high availability of fkbp5, gmeb1, and gmeb2 genes at a connection hub of temporo-limbic brain networks. Fkbp5 is known for having a role in stress-related psychiatric disorders, while gmeb1 and gmeb2 encode for modulator proteins of the glucocorticoid receptor, a key receptor in the hormonal stress system. Semantic networks tremendously ease working with multimodal neuroimaging and neurogenetics data and may reveal relevant coincidences between transcriptome and connectome networks.

脑连接组被视为健康与疾病状态下大脑功能的关键。结构性和功能性神经影像学技术使我们能够测量活体人类大脑的脑连接性。连接组学领域将连接组描述为一个数学图，其中连接强度由连接矩阵表示。图论算法被用于评估整个图的完整性，并揭示大脑网络生物标志物以用于大脑疾病的研究；然而，单个连接或子网络的结构异常作为神经或精神疾病的结构关联仍难以捉摸。我们提出了一种新颖的方法，通过语义网络来表示人类脑连接的知识——一种在知识管理中常用于描述对象间语义关系的形式化方法。在我们的新颖方法中，对象代表脑区，连接被建模为它们之间的语义关系。语义网络将连接组的图转化为关于哪些脑区相互连接的显式知识库。此外，此方法可以通过来自本体、脑图谱和分子生物学数据库的语义上下文，丰富个体受试者的测量连接性。将所有测量和事实整合到一个统一的特征空间中，使得跨模态比较和分析成为可能。我们利用语义网络的查询机制提取功能、结构和转录组网络。我们发现，一般来说，较高的结构和功能性连接性与连接脑区之间的较低差异基因表达相关；然而，皮层下运动区和边缘结构在强连接的同时，表现出局部化的高差异基因表达。在额外的探索性案例中，我们能够在额叶边缘脑网络的连接中心观察到fkbp5、gmeb1和gmeb2基因的高可用性。fkbp5因其与压力相关的精神疾病的作用而闻名，而gmeb1和gmeb2编码糖皮质激素受体的调节蛋白，糖皮质激素受体是激素压力系统中的关键受体。语义网络极大地简化了多模态神经影像学和神经遗传学数据的处理，并可能揭示转录组和连接组网络之间的相关巧合。