Analysis of host-pathogen interactions via clustering, statistical analysis, and data visualization

Abstract

Infectious diseases are caused by a variety of agents: viruses, bacteria, parasites, or even proteins. Using existing state-of-the-art methods and tools I developed myself, I studied aspects of infectious agents. To find the most conserved and diverse regions of influenza A proteins, I found clusters of extremely conserved or diverse residues. Because traditional methods of clustering proved ineffective for diverse regions, I developed a Metropolis Criterion Monte Carlo (MMC) clustering algorithm to discover clusters of extremely diverse regions. In addition to viruses, I studied pathogenic bacterial proteins known as effectors. Using an in-house prediction method, Preffector, I generated predicted effectors for 14 bacteria and created a database and webserver to hold relevant information: BacPaC. BacPaC uses intuitive visualizations and script-generated profile pages to display relevant data about the predicted effectors. Finally, I applied structural modeling and docking techniques to soybean proteins that are known to incur resistance to nematodes. For each of these studies, I used clustering, data analysis, and data visualization to better understand infectious agents.