- on the way to engineering the microbiome
The dynamics of the microbiome constituents can be modelled from longitudinal data. In its simplest form this can be modelled from the persistence probability (the probability that species will persist in a system over time) and the colonization probability (the probability that a species colonizes a system, given exposure). Importantly, these probabilities can be conditioned. For example, the probability that a bacteria persists in a microbiome may depend on the presence of a specific bacteriophage, another species, or some prebiotics (the condition). Such probabilities and their conditions can be learned from properly designed longitudinal microbiome studies. Notably, we have extensive experience with modeling persistence probabilities and their conditions in the human gut microbiome, and have expertise in designing Bayesian models for learning dynamics probabilities of complex microbial communities.
Figure: The survival probability of the mouse in this "experiment" may be conditioned by whether it is wearing the helmet or not