DECONVOLUTION OF CLONAL HETEROGENEITY FROM BULK AND SINGLE-CELL VARIATION DATA

项目摘要 随着单细胞基因组技术的成熟，细胞间的异质性是一种明显的 多细胞系统的普遍特征，对健康和疾病具有深远的影响。细胞到细胞 基因组异质性已成为包括组织在内的众多生物医学领域的研究热点 发育、再生、神经活动、传染病动力学和免疫反应。蜂窝 异质性在癌症中得到了最好的研究，在癌症中，它现在被理解为既常见又关键 肿瘤的生长、进展、转移和治疗反应。然而，尽管它有所有的希望，但也有 目前单细胞基因组学的基础研究和翻译应用之间存在巨大障碍。在……里面 特别是，单细胞基因组图谱的成本仍然高达数量级，无法达到所需的规模 在多组研究对象中描述细胞异质性，使其无法用于识别 患者队列中的异质性具有统计上的强健特征，更不用说常规的临床使用了。 真正的单细胞基因组学的一种更具成本效益的替代方案是基因组计算策略 去卷积，它使用来自组织样本的大量基因组测量来推断 在这些样本中共享的主要克隆。这种方法最近引起了人们对这两种方法的兴趣 计算生物学开发人员社区和主要癌症研究测序的分析管道 努力是描述肿瘤内异质性的一种经济有效的方法。它们有很大的缺点。 相对于真正的单元格数据，因为它们通常只提供少数几个非常不精确的重建 高丰度克隆种群。它们尤其不适合去卷积基因组中的信号 存在拷贝数变异(CNV)，这是大多数实体肿瘤进展的主要机制。 拟议的工作将力求解决对准确但成本效益高的研究方法的需求。 结合真正的单细胞基因组学和计算的优势实现细胞异质性 去卷积。这项工作将开发使用有限数量的单细胞数据的方法来增强 体积数据计算反卷积的精度和分辨率仅为真正单胞成本的一小部分 侧写。它将在两个主要变种中发展这一方向，特别关注癌症的背景：一个 将批量和单细胞测序(ScSeq)数据相结合，另一种将批量测序与 荧光原位杂交(FISH)，一种可以在有限的时间内表征克隆种群的技术 每个细胞的CNV标记物的数量，但细胞数量远远多于scSeq的实际数量。它将验证 由此产生的方法适用于散装、单细胞和FISH图谱均可获得的肿瘤样本。这个 由此产生的方法将提供一种方法，使单细胞基因组学今天在所需的规模上实用 对受试者队列和潜在的未来临床应用进行稳健的统计分析。

Project Summary As single-cell genomic technologies have matured, it has become apparent that cell-to-cell heterogeneity is a ubiquitous feature of multicellular systems with far-reaching consequences for health and disease. Cell-to-cell genomic heterogeneity has become a research focus in numerous biomedical contexts, including tissue development, regeneration, neural activity, infectious disease dynamics, and immune response. Cellular heterogeneity has been best studied in cancers, where it is now understood to be both common and crucial to tumor growth, progression, metastasis, and therapeutic response. For all its promise, however, there are substantial hurdles between current basic research and translational applications of single-cell genomics. In particular, single-cell genomic profiling is still orders of magnitude too costly to perform at the scales needed to profile cellular heterogeneity in more than small groups of study subjects, making it unusable for identifying statistically robust features of heterogeneity across patient cohorts, much less for routine clinical use. A more cost-effective alternative to true single-cell genomics is the computational strategy of genomic deconvolution, which uses bulk genomic measurements from tissue samples to infer the genomic signals of major clones shared among those samples. Such methods have seen a recent flurry of interest in both the computational biology developer community and in analysis pipelines of major cancer research sequencing efforts as a cost-effective way of profiling intratumor heterogeneity. They have substantial disadvantages relative to true single-cell data, though, as they provide generally only very imprecise reconstructions of a few high-abundance clonal populations. They are especially ill-suited to deconvolving genomic signals in the presence of copy number variations (CNVs), the major mechanism of progression in most solid tumors. The proposed work will seek to address the need for accurate but cost-effective methods for studying cellular heterogeneity by combining the advantages of true single-cell genomics with computational deconvolution. The work will develop methods for using limited amounts of single-cell data to enhance the accuracy and resolution of computational deconvolution of bulk data at a fraction of the cost of true single-cell profiling. It will develop this direction, with specific focus on the cancer context, in two major variants: one combining bulk and single-cell sequencing (scSeq) data and the other combining bulk sequencing with fluorescence in situ hybridization (FISH), a technology that can characterize clonal populations at limited numbers of CNV markers per cell but in far greater numbers of cells than is practical for scSeq. It will validate the resulting methods on tumor samples for which bulk, single-cell, and FISH profiles are all available. The resulting methods will provide a way to make single-cell genomics practical today on the scales needed for robust statistical analysis of subject cohorts and for potential future clinical applications.