Spatial Point Process Modeling of Viral Infections in Cell Cultures
Open Access
- Author:
- Simeonov, Ivan Botev
- Graduate Program:
- Statistics
- Degree:
- Doctor of Philosophy
- Document Type:
- Dissertation
- Date of Defense:
- June 12, 2012
- Committee Members:
- John Fricks, Dissertation Advisor/Co-Advisor
Francesca Chiaromonte, Committee Chair/Co-Chair
Murali Haran, Committee Member
Mary Poss, Special Member - Keywords:
- Spatial Point Processes
pair correlation function
in vitro cell culture studies
parametric point process models - Abstract:
- Biologists are often interested in investigating the progression of viral infections and co- infections (more than one virus, or more than one strain of the same virus) within a host organism. Various studies point to the fact that prior infections may substantially affect susceptibility to subsequent infections not just through responses due to the immune system, but through changes in the state of the cells exposed to infection within a tissue [Lo et al., 2005a; Martinez et al., 2007; Ruggiero et al., 1989]. In this thesis, I consider the response of epithelial cells to subsequent infections with two strains of the human Respiratory Syncytial Virus (A and B) in an in vitro experimental system. First, I use tools from exploratory spatial statistics and an appropriately devised simulation framework to investigate the likelihood of a cell being infected given its proximity to an infected cell [Simeonov et al., 2010]. Next, I focus on an exploratory summary statistic often used in the analysis of spatial point patterns called the pair correlation function. Its estimation is performed non-parametrically using kernel smoothing methods [Guan, 2007a,b; Guan et al., 2006] and bandwidth selection can substantially affect the estimation outcome. I develop a procedure for selecting the bandwidth that is effective and computationally viable also for large point patterns ( > 10000 points). I demonstrate the performance of our procedure through simulations and apply it to study the response of human epithelial cells to subsequent infections with human Respiratory Syncytial Virus strains A and B. Finally, I investigate parametric models to represent these data and inferential procedures to estimate their parameters. Elaborating on existing literature [Geyer,1999; Moller and Waagepetersen, 2003] I propose new models, prove that they satisfy certain stability conditions, and discuss their utility and various practical aspects of their parameter estimation via Markov chain maximum likelihood.