UC San Diego

Microbial communities play a crucial role in human health. The study of these host resident microbial communities is limited by the current inability to efficiently cultivate each member. Therefore, high throughput sequencing has been widely adopted to describe which microbial strains are present and in some cases their activity, directly from the sampled environment. The central aim of many microbiome studies is to understand how these communities form, perpetuate, and can be altered across time for the benefit of the host. However, several technical limitations in the analysis of these data have left many of these questions only partially revealed.First, in Chapter 1 of this thesis, we review the current understanding of the human microbiome across age, from birth until death. We also highlight many of the open gaps in our knowledge of how human microbiomes are formed and sustained, in addition to the current methodologies for exploring these questions. Chapter 2, introduces the current microbiome-based dimensionality reductions, which take high-dimensional microbiome data and reduces it into a few human interpretable dimensions. In this chapter we describe the limitations of these methods in microbiome data including sparsity, nonnormality, and compositionality. We address these problems with a novel method for dimensionality reduction which uniquely handles the inherent challenges of this data type. However, microbiomes are also highly individualized with each person containing a unique set of microbial communities. To overcome this, longitudinal studies of the microbiome are growing in popularity. In chapter 3, we describe the challenges to analyze these valid study designs, in addition to the lack of methods existing to properly account for the structure. In the chapter we address these challenges through tensor-based factorization, which accounts for the structure of the study. Through this method we are able to better understand microbial community development in infants, in particular those altered through C-section rather than vaginal birth. Finally, in chapter 4, we utilize these methods to explore a method for naturalizing C-section birth through seeding of the infant with the mother’s vaginal microbiome at birth. From this we found that engraftment of the mother’s vaginal microbiota at birth successfully naturalizes the microbiome development.