Fig. 1

MITRE learns human-interpretable rule-based models linking features of microbiota time-series data to host status. Rules operate on automatically learned time periods and groups of phylogenetically related microbes. a Schematic of the MITRE analysis pipeline, resulting in a single best predictive model as well as a distribution over alternative models that can be interactively explored. b Schematic of example rule in a applied to hypothetical data. Here, two subjects satisfy both the condition on the average abundance of microbe group A and the rate of change of abundance of group B