Center for Quantitative Biology: A focus on "omics", from organisms to single cells

单细胞分析的最新进展使数千到数十万个单个细胞的DNA序列、RNA表达、甲基化和蛋白质丰度的全基因组测量成为可能。捕获或推断空间或时间信息的方法提供了额外的上下文信息，以创建基因活动和功能的详细的细胞水平图像。尽管全基因组单细胞图谱为复杂组织的生物学提供了前所未有的见解，但由于大量经过测试的假设和随之而来的低统计能力和难以解释，在逐个基因的基础上分析这些数据是非常具有挑战性的。虽然基因水平的分析对大量组织数据提出了类似的问题，但由于噪声增加，零计数膨胀和多模态分布，单个细胞数据的问题被放大了。解决这些挑战的一个有希望的方法是基因集测试或途径分析。基因集测试最初是为大量组织数据开发的，是一种假设聚合方法，它利用有关基因之间功能关系的先验知识来测试较少数量的假设，从而提高解释、复制和统计能力。通过结合对一组或途径中所有基因的单细胞测量，基因集测试还可以减少方差并减轻稀缺性和多模型分布的影响。不幸的是，单细胞和大量基因组数据之间的统计和生物学差异使其无法在单细胞数据上开发用于大量组织的测试方法。尽管有定制技术的动机，但开发专门用于单细胞数据的基因集测试方法或集合的工作很少，对于一些重要的应用，如动态过程的分析，相关的方法还不存在。我们将通过开发一套针对单个细胞基因表达数据特征进行优化的基因集测试方法，解决当前支持单细胞数据基因集分析的局限性。为了验证这些方法，我们将研究组织驻留记忆CD8 T细胞的发育和功能。

Recent advances in single cell analysis have enabled the genome-wide measurement of DNA sequence, RNA expression, methylation and protein abundance for thousands to hundreds-of-thousands of individual cells. Methods that capture or infer spatial or temporal information provide additional contextual information to create a detailed, cell-level picture of gene activity and function. Although genome-wide single cell profiling provides unprecedented insights into the biology of complex tissues, analyzing such data on a gene-by- gene basis is extraordinarily challenging due to the large number of tested hypotheses and consequent low statistical power and difficult interpretation. While a gene-level analysis presents similar problems for bulk tissue data, the issues are magnified for single cell date due to increased noise, inflated zero counts and multi-modal distributions. One promising approach or addressing these challenges is gene set testing, or pathway analysis. Originally developed for bulk tissue data, gene set testing is a hypothesis aggregation method that leverages prior knowledge regarding the functional relationships between genes to test a smaller number hypotheses and thereby improve interpretation, replication and statistical power. By combining the single cell measurements for all genes in a set or pathway, gene set testing can also decrease variance and mitigate the impact of scarcity and multi-model distributions. Unfortunately, statistical and biological differences between single cell and bulk genomic data make it testing methods developed for bulk tissue on single cell data. Despite the motivation for customized techniques, little work has been done to develop gene set testing methods or collections that are specialized for single cell data and , for some important applications like the analysis of dynamic processes, relevant methods do not yet exist. We will address the limitations of current support for gene set analysis of single celldata by developing a suite of gene set testing methods optimized for the characteristics of single cellgene expression data. To validate these methods, we will us with the development and function of tissue-resident memory CD8 T cells.