Epidemiologic Studies of the Human Microbiome and of COVID-19

Abstract

Across these three chapters, we utilize and evaluate analytical and epidemiological methods for studying how humans and microbes interact, with a focus on two pathogens, Staphylococcus aureus and SARS-CoV-2. We first identified predictors of S. aureus carriage in a paired, longitudinal, high-dimensional dataset. We determined that while mothers represented an early source for S. aureus transmission to the developing infant microbiome, microbiome determinants became more important later on. We also identified a gene family that was significantly anti-correlated with S. aureus in infants and mothers and was likely acting as a phylogenetic marker for a bacterial species not closely homologous to current reference isolates that competes with S. aureus. Secondly, we evaluated methods for performing statistical mediation analysis with microbiome data in realistic synthetic data. We used our findings to perform mediation analyses exploring the role of the gut microbiome as a mediator between diet and cardiometabolic disease in two datasets. We identified the top performing methods for total indirect effect estimation, hypothesis testing for the total indirect effect, and hypothesis testing for component indirect effects. However, more work is needed to improve method performance and usability; currently, an end-user may need to employ multiple methods to accomplish a full mediation analysis with microbiome data, which can give rise to conflicting estimates. Finally, with the onset of the SARS-CoV-2 pandemic, we described how epidemiological biases, such as confounding, selection bias, and measurement error, can occur across five important classes of research questions for SARS-CoV-2 and COVID-19 and provided ways to avoid these biases across different stages of the study process.