Spatiotemporal Modeling of Microbial Communities
Microbial communities can undergo rapid changes, that can both cause and indicate host disease, rendering
longitudinal microbiome studies key for understanding microbiome-associated disorders. However, most
standard statistical methods, based on random samples, are not applicable for addressing the methodological
and statistical challenges associated with repeated, structured observations of a complex ecosystem.
Therefore, to elucidate how and why our microbiome varies in time, and whether these trajectories are
consistent across humans, we developed new methods for modeling the temporal and spatial dynamics of
microbial communities. We developed a method to identify ‘time-dependent’ microbes (Shenhav et al.,
PLoS Computational Biology 2019) and showed that their temporal patterns differentiate between the
developing microbial communities of infants and those of adults. We also developed models to deconvolute
the dynamics of microbial community formation. Using these methods, we found significant differences
between vaginally- and cesarean-delivered infants in terms of initial colonization and succession of their
gut microbial community (Shenhav et al., Nature Methods 2019) as well as the trajectories of these
communities in the first years of life (Martino*, Shenhav* et al., Nature Biotechnology). These models,
designed to identify and predict time-dependent patterns, will help researchers better understand the
temporal nature of the human microbiome from the time of its formation at birth and throughout life.