UC Irvine

Rapid growth of single-cell technologies provides unprecedented opportunities for close scrutinizing of heterogeneous cell states. However, detecting cell fate transition especially inferring the intermediate cell states (ICS) and transition cells from single-cell transcriptomic data remains challenging. In this dissertation, we focus on the epithelial-to-mesenchymal transition (EMT) as an example of cell fate transition. In Chapter 1, we introduce the existence and plausible biological roles of ICS in EMT. In Chapter 2, we present QuanTC, a method to infer cell fate transition, and a single-cell stochastic model of EMT which provides as a benchmark for QuanTC. In Chapter 3, we further apply QuanTC to single-cell transcriptomic datasets. We analyze the dynamical properties of inferred ICS based on a cell population model. In Chapter 4, we study the cellular crosstalk and the underlying gene regulatory dynamics along EMT from cancer cell lines with different inducing factors and find that the induced EMTs are context-specific. In Chapter 5, we combine deep learning with unbalanced optimal transport to model the temporal dynamics of time series single-cell transcriptomic data.