Leveraging Spatial Location for Single-Cell Molecular and Morphological Characterization

摘要 单细胞测序技术和技术的创新发展提供了 提高的分辨率和定义和表征细胞谱的新方法。尽管取得了这些进展， 细胞身份的不同方面，如转录组，空间位置，形态和生理 应对仍然具有挑战性。空间转录组技术，同时提供转录组数据内， 空间框架，经常必须妥协实现单细胞分辨率，以调查更广泛的面板 基因。类似地，虽然诸如荧光显微光学截面断层扫描（fMOST）和 功能超声成像（fUSI）提供了详细的神经元形态和生理重建， 然而，这些数据模式缺乏同时捕获分子信息的能力。因此，在本发明中， 虽然每种不同形式的技术进步不断解决更精细、更复杂的细胞类型 区别，结合联合收割机的细胞特性的所有方面，从转录组到 生理反应，还没有被捕获。因此，理解转录组学和 细胞的形态组成影响其生理反应是该领域的关键障碍。 该提案旨在开发将细胞身份的多个方面连接起来的计算工具。在 目的1，提出在积分非负矩阵分解中加入图正则化 算法（GRINMF）。使用GRINMF包括空间信息将导致更精细的细胞类型 用空间转录组学技术测定的细胞的表征。在目标2中，我们将验证空间 利用非负矩阵分解来计算细胞类型比例的去卷积算法 在空间配准的转录组数据中。我们将锚我们的衍生细胞类型比例体素在 与一系列形态和生理数据集相同的坐标框架。通过完成 提出的研究，我将获得丰富的经验，在算法的发展，以综合 生理学、转录组学和空间数据。这次培训将促进我的沟通， 批判性思维和翻译技能，为我提供了追求抱负所必需的工具 成为神经科学和生物信息学的研究科学家。

Abstract Innovative developments in single-cell sequencing technologies and techniques are providing increased resolution and novel ways to define and characterize cellular profiles. Despite this progress, linking different aspects of a cell’s identity, such as transcriptome, spatial location, morphology, and physiological response remains challenging. Spatial transcriptomic technologies, while providing transcriptomic data within a spatial framework, frequently must compromise achieving single-cell resolution in order to survey a wider panel of genes. Similarly, while techniques such as fluorescent micro-optical section tomography (fMOST) and functional ultrasound imaging (fUSI) provide detailed reconstructions of neuron morphology and physiological response, these data modalities lack the ability to simultaneously capture molecular information. As a result, while technological advances for each distinct modality continue to resolve finer and more complex cell type distinctions, cohesive cellular profiles that combine all aspects of a cell’s identity, from transcriptome to physiological response, have yet to be captured. Thus, understanding how the transcriptomic and morphological composition of a cell influences its physiological response is a key barrier for the field. This proposal aims to develop computational tools that will connect multiple facets of cellular identity. In Aim 1, we propose the addition of graph-regularization into the integrative non-negative matrix factorization algorithm (GRINMF). The use of GRINMF to include spatial information will result in more refined cell-type characterizations for cells assayed with spatial transcriptomics technologies. In Aim 2, we will validate a spatial deconvolution algorithm that leverages non-negative matrix factorization to calculate cell-type proportions within spatially registered transcriptomic data. We will anchor our derived cell-type proportion voxels in the same coordinate framework as a series of morphological and physiological datasets. By completing the proposed research, I will gain extensive experience in the development of algorithms to synthesize physiological, transcriptomic, and spatial data. This training will facilitate advancement of my communication, critical thinking, and translational technical skills, providing me with the tools necessary to pursue my ambition of becoming a research scientist at the interface of neuroscience and bioinformatics.