Mapping prediction error versus information content of coarse-grained descriptions of microbial ecosystems

Sequencing-based technologies allow resolving the composition of microbial ecosystems to strain-level detail; however, coarser representations are often found to be more reproducible and more predictive of community-level properties. The general principles for selecting an appropriate level of description for modeling remain elusive. We have recently developed a framework to begin addressing such challenges in the context of a standard ecological model of resource competition, where organisms are described functionally by a list of characteristics at an arbitrary level of detail. We build on this work to systematically compare all possible coarse-grained descriptions by explicitly quantifying their prediction power and information content. We show that coarse-grained descriptions can provide similar predictiveness as microscopic, but at a fraction of the entropy budget, allowing us to define an optimal "good-enough" description for predicting a property of interest. Finally, we demonstrate that the selected optimum can either increase or decrease in complexity as a function of community diversity, depending on the property we aim to predict. We discuss how investigating these behaviors nuances the notion of "emergent simplicity" in microbial ecology.