This project aims at identifying the dynamics of cellular phenotype switching from stochastic gene regulatory networks. Models for cellular phenotype switching enhance our mechanistic understanding of cell fate decisions, cellular differentiation processes and reprogramming, thus providing insight into, for example, tumor progression and cancer.
Nearly every cell in a person's body has the same DNA. However, thanks to gene regulation, each cell expresses (“turns on”) only a fraction of its genes at any given time. These different patterns of gene expression cause various cell types to have different sets of proteins, making each cell type uniquely specialized to do its job. In addition, gene expression is a fundamentally stochastic process. Randomness in transcription and translation leads to significant cell-to-cell variations in mRNA and protein levels, with important consequences for cellular function. Stochastic changes in particular patterns of gene expression can lead to spontaneous phenotype transitions (“switches”), being beneficial in some contexts and harmful in others. Stochastic gene regulatory networks can be modeled mathematically in terms of Markov jump processes and the Chemical Master Equation (CME). Given such a network model, we are interested in the number and characteristics of different phenotypes as well as in the transition probabilities and pathways between them. These numbers can be approximated from the CME by a projection method called Markov State Modelling. In this project, we want to overcome the current limitations of this method by developing new numerical algorithms and software tools that allow us to construct Markov State Models with quantified accuracy for high-dimensional stochastic gene-regulatory networks. We will demonstrate the relevance of Markov state modeling for understanding cellular phenotype switching by applying the algorithm to models for macrophage polarization, T-cell and stem cell differentiation.
This project aims at identifying the dynamics of cellular phenotype switching from stochastic gene regulatory networks. Models for cellular phenotype switching enhance our mechanistic understanding of cell fate decisions, cellular differentiation processes and reprogramming, thus providing insight into tumor progression and cancer. Cells are intrinsically noisy biochemical reactors, where low reactant numbers can lead to significant statistical fluctuations in molecule numbers and reaction rates. Such stochastic dynamical systems are modeled in terms of the Chemical Master Equation (CME), an infinite set of ordinary differential equations that is usually analytically intractable. Stochastic changes in particular patterns of gene expression have been identified with spontaneous phenotype transitions that can diversify otherwise identical cell-populations. These gene expression patterns correspond to preferred regions, called metastabilities, in the state space of the CME. The dynamics remains in one metastable region, before it rarely but rapidly switches to another metastable region. It has been shown in a proof-of-concept study that Markov state models provide a general framework for analyzing and visualizing metastabilities and state-transitions in gene networks described by the CME. However, the current approach requires an enumeration of the system state-space and the construction of a biochemical rate matrix and is therefore limited to systems with a small number of reactants. In order to make Markov state modelling a useful tool for quantitative systems biology at the network scale, we will adopt and extend the numerical algorithms to cope with high-dimensional systems. In particular, we will develop and implement a novel variational approach for a spectral approximation of the CME operator and apply this algorithm to the construction of Markov state models for multi-stable biological networks.