Advancing multiple model-based control of complex biological systems: Applications in T cell biology

<p>Activated CD4+ T cells are important regulators of the adaptive immune response against invading pathogens and cancerous host cells. The process of activation is mediated by the T cell receptor and a vast network of intracellular signal transduction pathways, which recognize and interpret antigenic signals to determine the cell's response. The critical role of these early signaling events in normal cell function and the pathogenesis of disease ultimately make them attractive therapeutic targets for numerous autoimmune diseases and cancers.</p> <p>Scientists increasingly rely on predictive mathematical models and control-theoretic tools to design effective strategies to manipulate cellular processes for the advancement of knowledge or therapeutic gain. However, the application of modern control theory to intracellular signal transduction is complicated by a unique set of intrinsic properties and technical limitations. These include complexities in the signaling network such as crosstalk, feedback and nonlinearity, and a dearth of rapid quantitative measurement techniques and specific and orthogonal modulators, the major consequences of which are uncertainty in the model representation and the prevention of real-time measurement feedback. Integrating such uncertainties and limitations into a control-theoretic approach under practical constraints represents an open challenge in controller design.</p> <p>The work presented in this dissertation addresses these challenges through the development of a computational methodology to aid in the design of experimental strategies to predictably manipulate intracellular signaling during the process of CD4+ T cell activation. This work achieves two main objectives: (1) the development of a generalized control-theoretic tool to effectively control uncertain nonlinear systems in the absence of real-time measurement feedback, and (2) the development and calibration of a predictive mathematical model (or collection of models) of CD4+ T cell activation to help derive experimental inputs to robustly force the system dynamics along prescribed trajectories. The crux of this strategy is the use of multiple data-supported models to inform the controller design. These models may represent alternative hypotheses for signaling mechanisms and give rise to distinct network topologies or kinetic rate scenarios and yet remain consistent with available data. Here, a novel adaptive weighting algorithm predicts variations in the models' predictive accuracy over the admissible input space to produce a more reliable compromise solution from multiple competing objectives, a result corroborated by several experimental studies. This dissertation provides a practical means to effectively utilize the collective predictive capacity of multiple prediction models to predictably and robustly direct CD4 + T cells to exhibit regulatory, helper and anergic T cell-like signaling profiles through pharmacological manipulations in the absence of measurement feedback. The framework and procedures developed herein are expected to widely applicable to a more general class of continuous dynamical systems for which real-time feedback is not readily available. Furthermore, the ability to predictably and precisely control biological systems could greatly advance how we study and interrogate such systems and aid in the development of novel therapeutic designs for the treatment of disease.</p>

Name thesis:degree_name

Doctor of Philosophy (PhD)

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Biomedical Engineering

Year

2016

Author dc:creator

• Perley, Jeffrey P

Contributors dc:contributor

• Ann E. Rundell
• Ed Bartlett
• Gregery Buzzard
• Marietta Harrison
• David Umulis

• Applied sciences
• Design of experiments
• Mathematical modeling
• Model predictive control
• Systems biology
• T cells
• Biomedical Engineering and Bioengineering

Repository record dc:identifier

https://docs.lib.purdue.edu/open_access_dissertations/695

OAI identifier oai:identifier

oai:docs.lib.purdue.edu:open_access_dissertations-1838