Analysis of the microbiome over time is hard. You can treat the abundance of each bacterial taxa as a separate outcome and look at them all individually but then you lose interactions between them. Worse yet, you can treat each measurement time point as separate, to get a picture of the interactions, but lose the connections between time points.
This week in journal club we will explore an alternative to longitudinal gut microbiome modeling – Network-based methods. After a simple introduction to longitudinal modeling and the different types of network methods we will look at a new implementation of one network method for longitudinal microbiome analysis from McGeachie et al. 2016 called Dynamic Bayesian Networks. In this paper they present the method and try it out on real data (16S rRNA gene profiles) from infants sampled every day in a neonatal intensive care unit. It’s important that you attempt the paper and google some of the terms before attending but I’ll do my best to explain the concepts used so that we can have a lively discussion.
Journal club will be this Friday April 29th from 3 – 4 pm in MUMC 3N10A. Afterward we will retire to the pub for a pint and to talk about a new view of the tree of life or maybe What Is the Tree of Life?
McGeachie MJ, Sordillo JE, Gibson T, Weinstock GM, Liu YY, Gold DR, Weiss ST, Litonjua A. (2016) Longitudinal Prediction of the Infant Gut Microbiome with Dynamic Bayesian Networks. Scientific Reports 6.