»Ê¼Ò»ªÈË

Abstract: Microbiomes are complex ecosystems wherein thousands of microorganisms spanning several kingdoms of life can coexist. Observed covariation in microbiomes suggests that their dimensionality may be substantially lower than the number of organisms. To quantify this dimensionality, we developed a latent variable generalization of the consumer/resource (C/R) model with a tunable dimension. Fitting this model to several datasets showed that dimensionality varied substantially across communities with many microbiomes detected to be very low dimensional. Low dimensionality of microbiomes can arise either from correlations in species resource preferences or correlations in resource inflows. Simulations using the C/R model showed that resource inflows but not resource preferences were responsible for the observed low dimensionality. The model predicted that low dimensional microbiomes had enhanced competition amongst species. Metabolic modeling confirmed these predictions in real microbiomes. Additionally, low dimensional microbiomes accurately predicted the state of their environment. We used this insight to develop a generative model for rational microbiome engineering. In the model, variation in ecosystem composition arose due to differences in effective resources (latent variables) while species’ resource preferences remained conserved. The latent variables also explained phenotypic states of hosts.  The model allowed us to address several salient questions in microbial ecosystems: (1) which environmental factors maximally constrain the composition of microbiomes? (2) how context-specific are phenotype/microbiome associations, and (3) what are plausible microbiome compositions that correspond to desired host phenotypes? Therefore, mechanistic models are not only able to shed qualitative light on organization of complex microbial ecosystems but when appropriately modified can also allow us to design these ecosystems.

Host: Kabir Haussain(kabir.husain@ucl.ac.uk)