Single-cell gene expression dynamics during neurogenesis

项目摘要/摘要 神经外胚层、神经脊和神经前体细胞群体的适当规格决定了 神经发育的总体成功，以及这些群体的破坏是许多发育的根源 疾病。单细胞转录学的最新进展使数万人能够研究这一过程 整个转录组中的细胞。与批量测序不同，批量测序是跨群体平均表达的，单个 细胞方法对于研究像神经发育这样的多方面、不同步的过程是必不可少的。 然而，分析scRNA-seq数据存在几个计算障碍，包括辍学、技术噪声、 和批处理效果。此外，研究发育中的神经组织本身也存在挑战，因为在许多情况下 上下文、祖细胞和成熟后代同时存在。因此，从生物角度获得洞察力 这些数据需要开发新的计算方法。这项提案的总体目标是建立 研究神经发育过程中单细胞基因表达的计算方法，重点是人类 类胚体(EBS)。在目标1中，我将描述胚层的分支结构和神经规范 在27天的EB培养过程中。在我的初步结果中，我使用了PHATE，这是一种降维工具 共同开发的，描述了一套从干细胞通过三个胚层到几个胚层的平稳过渡 衍生物包括心脏、骨骼和神经谱系的祖细胞。我将验证这个结构 预测神经外胚层和神经前体细胞的FACS分选和批量RNA测序鉴别 细胞。在目标2中，我将描述基因表达的平滑变化-神经的驱动规范 通过推断EB时间进程中的潜在发育时间来推断血统。为每个单元格分配唯一的单元格 发展时间标签，我开发了一种名为MELD(潜在维度流形增强)的方法。在……里面 在我的前期工作中，我使用MELD来概括已知的基因表达模式 神经脊细胞。在这个目标中，我将把我的分析扩展到神经外胚层和神经前体种群。 并确认MELD预测在未在原始中采样的时间点期间出现的中间小区状态 时间航向。这一目标的工作将产生一个连续的路线图，从干细胞到 EBS中的神经前体。在第三个目标中，我将使用互信息来研究转录的边缘重连 因素及其目标(在神经发生过程中调节关系如何变化)。通过对 监管因素及其目标之间的统计相关性作为一个连续的过程，将是可能的 推断调控因子表达对其影响最大的关键窗口 下游目标。为了确认这些窗口的时间，我将使用多西环素诱导来显示诱导 在窗口内对下游转录的影响比在窗口外更大。总而言之，这些见解将 从单细胞RNA测序数据中产生更好地了解神经发生的方法。

PROJECT SUMMARY/ABSTRACT The proper specification of the neuroectoderm, neural crest, and neural progenitor populations dictates the overall success of neurodevelopment, and disruption of these populations is at the root of many developmental diseases. Recent advances in single cell transcriptomics allow for the study of this process in tens of thousands of cells across the transcriptome. Unlike bulk sequencing, which averages expression across a population, single cell approaches are essential to studying a multifaceted, asynchronous process like neurodevelopment. However, several computational barriers exist to analyzing scRNA-seq data, including dropout, technical noise, and batch effects. Furthermore, studying developing neural tissue provides its own challenges because in many contexts, progenitor cells and mature progeny are present at the same time. Thus, gaining biological insight from these data requires development of novel computational methods. The overall goal of this proposal is to establish computational methods to study single cell gene expression during neurodevelopment, focusing on human embryoid bodies (EBs). In aim 1, I will characterize the branching structure of germ layer and neural specification in a 27-day time course of EB culture. In my preliminary results, I used PHATE, a dimensionality reduction tool I co-developed, to describe a set of smooth transitions from stem cell through the three germ layers to several derivatives including progenitors of the cardiac, bone, and neural lineages. I will validate this structure of differentiation by FACS sorting and bulk RNA-sequencing of predicted neuroectoderm and neural progenitor cells. In aim 2, I will describe the smooth changes of gene expression the drive specification of the neural lineages by inferring the latent developmental time in the EB time course. To assign each cell a unique single developmental time label, I developed a method called MELD (Manifold Enhancement of Latent Dimensions). In my preliminary work, I use MELD to recapitulate known patterns of gene expression during the specification of neural crest cells. In this aim, I will expand my analysis to the neuroectoderm and neural progenitor populations and confirm that MELD predicts intermediate cell states present during time points not sampled in the original time course. The work in this aim will generate a continuous roadmap of gene expression from stem cell through neural progenitor in EBs. In the third aim I will use mutual information to study edge-rewiring of transcription factors and their targets (how the regulatory relationships change during neurogenesis). By modelling the statistical dependency between regulatory factors and their targets as a continuous process, it will be possible to infer the crucial windows during which regulatory factors expression has the strongest influence on downstream targets. To confirm the timing of these windows, I will use doxycycline induction to show induction has a greater effect on downstream transcription within windows than outside them. Together, these insights will produce methods to gain a greater understanding of neurogenesis from single cell RNA-sequencing data.