Data-driven stochastic modelling of gene regulation
File(s)
Author(s)
Volteras, Dimitris
Type
Thesis
Abstract
Gene expression varies significantly from cell to cell and over time, giving rise to phenotypic diversity within populations of genetically identical cells. This variation has a crucial role in gene regulatory networks and cell fate decisions. The recent rise in single-cell genomics technologies promises to exploit cellular variability to elucidate mechanisms of gene regulation. However, little emphasis is given on developing stochastic models that can effectively harness single-cell data to make sense of this variability. Moreover, single-cell sequencing data only reveals static snapshots of gene expression and is corrupted by technical noise, hindering biologically meaningful variation and posing challenges in uncovering gene regulatory mechanisms. The first part of this thesis introduces a stochastic modelling framework that leverages time-resolved single-cell transcriptomics data to tackle these challenges. The method delineates biological and technical noise sources and employs Bayesian inference to reveal global mechanisms of transcriptional regulation. The second part extends the framework to analyse gene co-expression relationships in time-resolved single-cell data, using stochastic models of gene co-regulation. The approach quantifies biological and technical contributions to transcriptional covariation and utilises machine learning techniques to identify modes of gene interaction. The final part attempts to address challenges arising from the complexity of modelling gene regulatory networks. It presents a rigorous theoretical method for approximating stochastic biochemical networks under timescale separation conditions, with applications to models of gene regulation.
Version
Open Access
Date Issued
2024-09-03
Date Awarded
01/03/2025
License URL
Advisor
Thomas, Philipp
Shahrezaei, Vahid
Sponsor
Imperial College London
Publisher Department
Department of Mathematics
Publisher Institution
Imperial College London
Qualification Level
Doctoral
Qualification Name
Doctor of Philosophy (PhD)