Modeling the Gene Regulatory Dynamics in Neural Differentiation with Single Cell Data Using a Machine Learning Approach
Author | : Yixing Hu |
Publisher | : |
Total Pages | : |
Release | : 2022 |
ISBN-10 | : OCLC:1310301248 |
ISBN-13 | : |
Rating | : 4/5 ( Downloads) |
Download or read book Modeling the Gene Regulatory Dynamics in Neural Differentiation with Single Cell Data Using a Machine Learning Approach written by Yixing Hu and published by . This book was released on 2022 with total page pages. Available in PDF, EPUB and Kindle. Book excerpt: "Cellular differentiation is an important process where progenitor cells progressively develop into mature cells with specialized functions. Understanding the molecular characteristics and underlying regulatory mechanisms of cell fate is a central goal in biological research. Advances in single-cell sequencing technology enable the exploration of cellular differentiation at unprecedented resolution. In this thesis, I focus on characterizing and modeling of cellular differentiation using machine learning approaches. First, I present a random forest approach to identify the most discriminant genes for different cell populations in the developing brain. This method was able to identify key gene markers that revealed dorsal-ventral patterning in a heterogeneous class of progenitors present in a mouse developmental time-series dataset. Next, as cellular differentiation is marked by continuous changes in gene expression and is not well described by static cell populations, I present a framework to model the dynamics of cell fate decisions based on ordinary differential equations (ODE). I train this model on previously reported trajectory data for neural differentiation, and show that the model is able to interpolate and predict the gene expression dynamics across unobserved regions in this trajectory and extend the system dynamics for neural differentiation data. Finally, by training the model on datasets that contain rate of change information for each gene (RNA velocity), I demonstrate that the model has the capacity to predict the effects of gene deletions to the cell's overall gene expression profile with a prediction accuracy of 90%. In summary, the Neural ODE method has the ability to learn the gene regulatory dynamics from single cell data and predict the dynamics of individual genes as well as perturbation response"--