Attractor identification and control of Boolean and ODE network models in systems biology
Open Access
- Author:
- Rozum, Jordan
- Graduate Program:
- Physics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- May 11, 2022
- Committee Members:
- Dezhe Jin, Major Field Member
Lu Bai, Major Field Member
Reka Albert, Chair & Dissertation Advisor
Timothy Reluga, Outside Unit & Field Member
Nitin Samarth, Program Head/Chair - Keywords:
- systems biology
network science
discrete dynamics
Boolean networks
gene regulatory networks
ordinary differential equations
control theory
stable motifs - Abstract:
- Cells use chemistry to process complex information and make decisions. How should a stem cell differentiate? What prompts a cancer cell to enter a metastatic state? How does a cell know when to undergo apoptosis? Over the past several decades, a growing trove of biomolecular data has enabled detailed insights into the gene and protein interactions that underlie these biomolecular decisions. Still, much work remains to transform this unprecedented wealth of data into cohesive understanding and to leverage that understanding to efficiently develop novel clinical therapies. A crucial step toward these goals is the construction and analysis of predictive dynamical models, which integrate biomolecular data to generate testable and mathematically precise predictions about the effect of genetic modifications and pharmaceutical interventions. These models are essential for evaluating biological understanding of the complex behavior that emerges from biomolecular circuits, and they have long-term applications in personalized medicine and automated drug target identification. A fundamental difficulty that limits modeling in both clinical and laboratory settings is that, due to their high dimension and extreme nonlinearity, these models are computationally difficult to construct and analyze. One approach, first developed for the analysis of qualitative discrete models of biomolecular circuits, is to analyze so-called “stable motifs”, or self-sustaining patterns of activity in small subcircuits within a larger model. Collective dynamics of the entire network structure is inferred from the interactions between these patterns and their downstream effects. In this dissertation, I discuss three areas of research in which I have applied the stable motif concept across six published articles to advance the state-of-the-art in both qualitative and quantitative biomolecular model development and analysis. In Chapter 1, I review the traditional stable motif approach to analyzing Boolean network models of biomolecular circuits. In Chapter 2, I present my work on the extension of stable motifs to study oscillations in asynchronously updated Boolean networks and an application to a model of the genetic circuitry that drives the cell cycle. In Chapter 3, I present my work on pystablemotifs, a Python library that implements efficient algorithms for the attractor identification in asynchronous Boolean networks. I showcase several applications of this software to empirical models of processes of critical importance in cancer development and proliferation. In addition, I present an application of pystablemotifs to answer a 50-year-old question about the prevalence of attractors in a widely studied phase transition in random Boolean networks. In Chapter 4, I present various control algorithms I have developed and implemented in pystablemotifs. These algorithms uncover overrides in the network of biomolecular interactions that can drive the system to a desired attractor from any initial state. In the final four chapters, shift focus from qualitative Boolean models to quantitative ordinary differential equation (ODE) models. In Chapter 5, I describe the unique challenges of applying control theory to biomolecular ODE models. I emphasize the benefits of control theoretic tools for identifying nonperturbative control interventions in high-dimensional systems that respect a specified set of privileged coordinates. I also discuss several traditional techniques and describe several techniques that make use of feedback loops in the network of interactions to identify node overrides that achieve desired attractor states. I developed one such method by generalizing the Boolean stable motif concept to a class of ODEs that are commonly used to model biomolecular circuits. I describe this generalization in Chapter 6 and provide example applications. Further applications to generic biomolecular circuits are described in Chapter 7. In Chapter 8, I describe how this method can be applied to help parameterize quantitative ODE models when a qualitative Boolean model of the same system is available. Finally, Chapter 9 presents conclusions and possible future applications of my PhD research.