Abstract 1231: A stochastic data-driven model predicts the timing and distribution of the epithelial-mesenchymal transition
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
AbstractThe Epithelial-Mesenchymal Transition (EMT) is a key contributor to cancer metastasis and morbidity. During this process, epithelial cells acquire migratory and stem-like characteristics, which allow them to travel through vasculature and form distant metastases in a secondary organ site via the reverse process of MET. Initially thought of as a binary process, it is now well established that cells can be stably arrested en route to EMT in an intermediate hybrid state (H) with characteristics falling on a spectrum between the epithelial (E) and mesenchymal (M) phenotypes. This non-transient intermediate state is stabilized by phenotypic stability factors and is associated with more tumor aggressiveness, worse treatment response, and patient outcomes. Although abundant time-course single-cell RNA sequencing data exists, enabling dynamical single-cell analysis of EMT, current methods of EMT state inference are optimized on microarray data. Thus, there is a large need for a computational framework that infers and predicts the E, H, and M fractions at transient and steady states from single-cell RNA sequencing data. Here, we introduce a fully stochastic continuous-time Markov model to predict the dynamics of EMT transitions from time-course data. We fit our continuous-time model on a multi-replicate time-course flow cytometry data of MCF10A cell line where TGF is withdrawn at different days following treatment for each replicate. We then demonstrate that our model is extendable to time-course single-cell RNA sequencing data utilizing a novel data-driven pipeline. Our data-driven pipeline infers the three E, H, and M trajectories from time-dependent and dose-dependent single-cell RNA sequencing data. We confirm our trajectories accurately reflect the E, H, and M states via gene-set enrichment analysis. We find the fraction of cells in each state at stationary distribution using our data-driven pipeline and show that they match the predictions of a previous simulation study. By applying the stochastic model to time-course single-cell RNA sequencing trajectories of four cancer cell lines (A549, MCF7, OVCA420, and DU145), we show that the stochastic model accurately predicts context-specific EMT dynamics over time. Our model is the first RNA sequencing-based method for tracking EMT from longitudinal data at the single-cell resolution and therefore is of immediate utility for studying metastatic signatures in cancer.Citation Format: Annice Najafi, Mohit K. Jolly, Jason T. George. A stochastic data-driven model predicts the timing and distribution of the epithelial-mesenchymal transition [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 1231.
