Our previous speaker emphasized how difficult microbial community interactions are to predict. Will Harcombe, from the University of Minnesota, wants to do just that. He and his colleagues are using computational approaches to link genomics and metabolic networks to predict community properties. Starting with a genome, using it to understand physiology, then linking it to species interactions, then to community properties – and then understanding how evolution and selection affect each of these steps – this encompasses the full range of questions Harcombe is interested in explaining. Using a computational tool called COMETS (Computation of Microbial Ecosystems in Time and Space), Harcombe and his colleagues were able to predict variation in colony size and accurately predicts species interactions from a synthetic consortia – not just the duration to equilibrium, but the ratios of species present.
Harcombe recommends using COMETS as a null model for predicting microbial community properties – COMETS can make a simple prediction, he says, and if it’s not accurate, there’s some other really cool biology going on that needs to investigated. What is the scale over which bacteria compete? Harcombe initially struggled to predict the behavior of microbial colonies in competition, but then he walked down the hall to talk to some physicists. They told him he needed to use territory diagrams, mapping areas according to which colony they are closest to. With these diagrams in his model, he was able to explain 99% of variation, showing that distance isn’t the only thing that matters to competing colonies, but geometry and territory size. COMETS also makes it possible to start linking genes (and mutations) to ecosystem properties. Knocking out genes one by one, Harcombe was able to simulate the effects of evolution and mutation at work in the community. For those seeking to predict community function and evolution from metagenomics and metabolic networks, COMETS serves as a potential tool for doing so.