Microbial community dynamics reduce to low dimensions that scale linearly with observables, study finds
The smallest latent dimension needed to reconstruct observable microbial dynamics is much lower than the community's apparent degrees of freedom, so some predictions need no full measurement.
Researchers reported in the Proceedings of the National Academy of Sciences that the temporal dynamics of observable microbial populations can be captured by low-dimensional representations pnas.org
Proceedings of the National Academy of Sciencespnas.orgOpen source ↗. Microbial communities show complex dynamics driven by many interactions among members and with their environments, yet modeling and control usually rely on a limited number of observables.
Critical latent dimension
Using variational autoencoders, the authors defined a critical latent dimension, Ec. It operationally quantifies the smallest latent dimension needed for high-fidelity reconstruction of observable microbial dynamics under a specified criterion. Despite the complexity of unobserved background dynamics, Ec was found to be much lower than the community's apparent degrees of freedom.
Ec scales linearly with the number of observables, which the authors present as an empirical principle for determining the complexity of a representation.
Where the scaling holds
The principle held across simulations of pairwise interacting communities, consumer–resource dynamics, spatially distributed microbial patches and plasmid transfer. It also held in experiments with engineered and environment-derived communities and in human microbiomes. pnas.org
Proceedings of the National Academy of Sciencespnas.orgOpen source ↗
According to the authors, these results establish a scaling law for microbial community dynamics. Low-dimensional representations of observable dynamics retained information useful for selected prediction tasks without direct measurement of the complete community.
