Gurdip Uppal, PhD
Gurdip Uppal, Georg K. Gerber
Georg K. Gerber
Research Category: Digital Health, Imaging, and Informatics
The microbiome, or collection of commensal micro-organisms that live on and within us, is extremely complex and plays key roles in many prevalent human diseases. Spatial structuring of the microbiome affects interactions and underlying community functions, which are critical to understand for purposes of altering the microbiome to treat and prevent diseases. However, little is known about the in vivo biogeography of the microbiome. Metagenomic plot sampling by sequencing (MaPS-seq) data is a novel culture-independent method that can characterize the spatial organization of an entire microbiome at micron-scale resolution, but produces data that is noisy and can be difficult to interpret. We present a novel computational method, based on embedded topic models, that addresses the challenges of analyzing MaPS-seq data, producing interpretable probabilistic maps of spatial organization among microbes and recovering distinct community subtypes. To solve the difficult inference task of combining spatial structure and community subtypes, we developed a new Markov Chain Monte Carlo-based inference algorithm that uses Polya-Gamma auxiliary variables. We demonstrate our method’s ability to provide new insights into the spatial organization of the microbiome along the murine intestinal tract, and how dietary changes alter spatial organization in the lower gastrointestinal tract.
Our microbiomes, or communities of trillions of beneficial micro-organisms that live on and within us, have been shown to play key roles in human health and disease. These microbial communities exhibit intricate spatial organization, which influences how the microbes interact with each other and with us, their host. Understanding this spatial organization is an important step to unlocking the potential of the microbiome to improve human health, because it can allow us, for instance, to predict which microbes can effectively colonize our bodies and prevent pathogenic bacteria such as C. difficile from infecting us. However, little is known about the spatial organization of the microbiome. In this work, we present a new computer-based method that analyzes data from an experimental method that measures detailed information about the spatial structure of microbiomes. Our method addresses challenges in analyzing this data, which is noisy and can be difficult to interpret, by automatically deducing the spatial relationships between microbes and finding sub-communities of microbes. We demonstrate our method’s ability to provide new insights into the spatial organization of the microbiome along the intestinal tract, and how dietary changes alter spatial organization in the lower gastrointestinal tract.