Technological developments in the past thirty years have transformed sequencing-based microbiology into a data-intensive field. Here, computing and efficient representations are catalyzers of insight into omnipresent and complex microbial interactions. Notably, classical ecologists have set the foundations for the way we analyze these systems, with some techniques dating back to the beginning of the twentieth century. In this thesis, we expand and where possible reuse these techniques to unravel the hidden patterns comprising the human gut microbiome.
To set an appropriate motivation and context for the rest of this work, Chapter 1 reviews recent discoveries on the human microbiome and how the communities within can influence the effectiveness of therapeutic agents. Next, in Chapter 2, we introduce EMPeror, an interactive analysis and visualization tool that is crucial to the findings presented in later chapters.
The following three chapters study concrete examples where the microbiome has been implicated as a driver or marker for dysbiosis. Chapter 3 describes how the microbial signature associated with Crohn's disease (CD) in humans, described in our previous work, is overlapping but distinct to that of dogs affected with inflammatory bowel disease (IBD). Surprisingly, unlike with humans, dog fecal samples alone are strong indicators of the disease. In Chapter 4, we study IBD from a longitudinal perspective, revealing increased volatility in the gut microbiomes of subjects with IBD, a property that does not appear to be present in unaffected controls. Furthermore, we use this as a predicting feature of the disease, and improve on the classification accuracy possible through a single fecal sample. In Chapter 5, we study the effect of fecal microbiota transplants (FMTs) to treat Clostridium difficile infection (CDI) and, using the techniques described in Chapter 2, we show the first animated visualization of this process, a dramatic microbial transformation as the subjects recover from all CDI symptoms. In addition, for CDI patients who also suffer from a subtype of IBD, a treatment with a FMT results in an increased number of relapses and decreased microbial diversity.
The closing chapter discusses these results and their possible applications, as well as future directions for computationally-centric microbiome research.