Inferring Developmental Trajectories and Causal Regulations with Single-cell Genomics
Development is commonly regarded as a hierarchical branching process which is governed by underlying gene regulatory networks. Single-cell genomics, single-cell RNA-seq (scRNA-seq) in particular, holds the promise to resolve the dynamics of this process. However, learning the structure of complex single-cell trajectories with multiple branches remains a challenging computational problem. In this thesis, I will present the toolkit, Monocle 2, which uses reversed graph embedding to reconstruct single-cell trajectories in a fully unsupervised manner. Monocle 2 learns an explicit “principal graph” that passes through the middle of the data as opposed to other ad hoc methods, greatly improving the robustness and accuracy of its trajectories. I will demonstrate that Monocle 2 is able to accurately reconstruct developmental trajectories for complicated systems, including hematopoiesis involving multiple different cell fates. When coupled with another statistical framework, BEAM (branch expression analysis modeling), Monocle 2 is able to detect genes specific to different developmental lineages. The unprecedented high resolution of the reconstructed developmental trajectories not only enables us to determine which genes are playing important roles at the critical time point of cell fate transition but also to directly infer causal gene regulatory networks. To this end, I have been developing a new toolkit, Scribe, which applies novel information theory techniques to detect causal interactions responsible for fate transitions. Scribe provides intuitive visualizations of causal interactions and can additionally incorporate information from “RNA-velocity” for causality detection. Scribe accurately reconstructs core networks specifying myelocytic or chromaffin cells. Finally, I will show a compendium of the inferred causal regulatory network for C elegans’ early embryogenesis based on lineage resolved live imaging data, demonstrating Scribe’s generalizability.
用单细胞基因组学推断发育轨迹和因果规则
发育通常被认为是由基础基因调控网络控制的分层分支过程。
单细胞基因组学，特别是单细胞RNA-seq（scRNA-seq），有望解决这一过程的动态变化。
然而，学习具有多个分支的复杂单细胞轨迹的结构仍然是具有挑战性的计算问题。
在本文中，我将介绍工具包Monocle 2，它使用反向图嵌入以完全无监督的方式重建单细胞轨迹。
Monocle 2学习了一个明确的“主图”，它通过数据的中间而不是其他特殊方法，大大提高了其轨迹的鲁棒性和准确性。
我将证明Monocle 2能够准确地重建复杂系统的发育轨迹，包括涉及多种不同细胞命运的造血。当与另一个统计框架BEAM（分支表达分析建模）结合时，Monocle 2能够检测特定于不同发育谱系的基因。
重建发育轨迹的前所未有的高分辨率不仅使我们能够确定哪些基因在细胞命运转变的关键时间点起重要作用，而且还能直接推断因果基因调控网络。
为此，我一直在开发一个新的工具包Scribe，它使用新颖的信息理论技术来检测导致命运转换的因果交互。
Scribe提供了因果关系的直观可视化，并且可以另外包含来自“RNA-速度”的信息以用于因果关系检测。 Scribe准确地重建了指定髓细胞或嗜铬细胞的核心网络。
最后，我将根据谱系解析的实时成像数据，展示推断的C线虫早期胚胎发育因果调控网络的概要，展示Scribe的普遍性。