Can trade-offs describe the coarse-grained structures in microbial communities?

One long-standing problem is the rise of diversity in microbiomes due to community structures. Recent experiments show that a small number of growth-limiting nutrients can sustain richly diverse systems. While their compositions on the resolution of strains are highly variable across samples, they converge at the coarse-grained level at higher taxonomic ranks, suggesting an underlying low-dimensional structure. For example, in laboratory experiments where abundances of clusters depended on specific nutrient sources, strains in the same clusters had similar metabolic capacity. In contrast, it has been shown that strain-specific interaction is crucial to capture the dynamics of the system due to their differences in metabolic functions. This opens up possibilities of models where clustering departs from similarities in phenotypes. Our work shows that strains exhibiting different strategies but sharing the same trade-offs can be clustered into the same group. Correlation of strain abundances generated from our model shows underlying hierarchical structures, which allows large systems (~100 strains) to be cast into small systems with only a handful of competing coarse-grained units. Similar to experiments, the clusters remain stable, while strains within each cluster show wide fluctuations.