Computational approaches for the systematic detection of cell-cell interactions by spatial transcriptomics - Resubmission - 1

摘要 许多生物过程不是在细胞水平上发生的，而是在系统水平上发生的，细胞与细胞之间的相互作用 对组织功能至关重要。随着单细胞RNA-Seq的引入，我们有了强大的细胞测量手段 类型和单元格状态。然而，在这种方法中，研究中的组织必须在测序之前分离 导致了空间语境的丧失。空间转录学是一个很有前途的新领域，其中有几个 已经开发出方法来描述细胞在其自然环境中的转录组。然而，最多的 这种技术的广泛应用--基于测序的空间转录组学--尚未达到 单元格分辨率。因此，迫切需要集成空间数据的新型计算方法 转录学和单细胞RNA-Seq，以推断复杂组织中细胞与细胞的关系。我们的实验室有 最近开发的对这两个数据源的多模式交集的分析有效地缓解了 每种技术的局限性。在这里，我们建议应用这个概念来揭示细胞-细胞的模式 生物系统中的相互作用。在我们的第一个目标中，我们提出了STATEMAP方法来推断局部细胞-细胞 基于空间转录的共定位和受体-配体关系的相互作用。单元格目录 首先使用单细胞数据来描述类型和细胞状态，然后是空间转录数据 被用来映射共定位细胞状态对。然后，STATEMAP系统地推断细胞与细胞的相互作用 通过统计测试信号/响应关系在共定位细胞状态之间的机制 空间转录学数据。在第二个目标中，我们提出了ST-Motif方法来概念化 作为网络的小区类型和状态的位置，允许通过丰富的可用的 方法：研究方法。因此，我们的方法将寻找细胞-细胞关系的问题重新定义为网络主题问题 在这张图中。在我们的整个提案中，我们在两个模型系统上开发和测试算法，Male 生殖系和胎盘，我们的实验室在这方面有丰富的经验。从概念上讲，我们的建议 有望产生新的算法来映射激活潜力所需的细胞-细胞相互作用 两种强大的转录技术。

SUMMARY Many biological processes occur not at the level of a cell but at the level of a system, and cell-cell interactions are crucial for tissue function. With the introduction of single-cell RNA-Seq, we have robust measures of cell types and cell states. In this approach however, the tissue under study must be dissociated prior to sequencing resulting in the loss of spatial context. Spatial transcriptomics is a promising new field, in which several methods have been developed to profile the transcriptome of cells in their native context. However, the most widely used implementation of this technology – sequencing-based spatial transcriptomics – has not reached single-cell resolution. Thus, there is a critical need for novel computational approaches integrating spatial transcriptomics and single-cell RNA-Seq in order to infer cell-cell relationships in complex tissues. Our lab has recently developed analyses for multimodal intersection of these two data sources that effectively mitigate the limitations of each technology. Here, we propose to apply this concept to uncover patterns of cell-cell interactions in biological systems. In our first Aim, we present the StateMap approach to infer local cell-cell interactions by spatial transcriptomics-based co-localization and receptor-ligand relationships. A catalog of cell types and cell states is first delineated using single-cell data, and the spatial transcriptomics data is then harnessed to map pairs of co-localizing cell states. StateMap then systematically infers the cell-cell interaction mechanisms among co-localizing cell states by statistically testing for signal/response relationships in the spatial transcriptomics data. In our second Aim, we propose the ST-motif method to conceptualize the locations of cell types and states as a network, allowing for systematic analysis by a wealth of available methods. Our approach thus reframes the problem of finding cell-cell relationships as a network motif problem in this graph. Throughout our proposal, we develop and test the algorithms on two model systems, the male germline and the placenta, with which our lab has considerable experience. Conceptually, our proposal promises to yield novel algorithms for mapping cell-cell interactions that are required for actuating the potential of two powerful transcriptomic technologies.