UCSF

Diseases related to the human gut microbiome, such as irritable bowel disease (IBD), irritable bowel syndrome (IBS), and colorectal cancer, have increased within developed nations. At the same time, increased access to metagenomics has allowed us to try to understand the microbiome. While many microbiome-disease association studies have been carried out, they primarily focus on genera or species that vary in abundance with disease status; this relatively coarse level of analysis limits our understanding of why microbes may act as disease markers and overlooks cases where disease risk is related to the presence or absence of specific strains with unique biological functions. My thesis shows that there are strong sub-strain phenotypic signatures across samples and then introduces microSLAM, a method that incorporates random effects to represent the population structure of the bacteria modeled in an association study. I applied microSLAM to many gut metagenomes from IBD samples, where I discovered 49 species whose population structure correlates with IBD. In addition, after controlling for population structure, I found 57 microbial genes that are significantly more common in healthy individuals and 26 that are more common in IBD patients, including a seven-gene operon in Faecalibacterium prausnitzii that is involved in the utilization of fructoselysine from the gut environment. In addition, I performed extensive simulations to understand the limitations and capabilities of microSLAM and found it was much more conservative and specific than standard a generalized linear model (GLM). These findings have highlighted the importance of considering within-species population structure in microbiome studies.