Abstract
How might we artificially select multi-species microbial communities to improve their functions? In the accompanying article, we have modeled a commensal community where Helper releases Byproduct essential to Manufacturer, and Manufacturer diverts a fraction fP of its growth to make Product. Low-density “Newborn communities” grow and mutate over time T into “Adult communities”, and Adult communities with the highest function (total Product) are randomly partitioned into Newborn communities of the next cycle. Here, to understand selection dynamics, we visualize community function landscape (“landscape”) which relates community phenotype composition (e.g. species ratio; Manufacturer’s fP) to community function. We show that although an interaction can enable species coexistence by driving species ratio toward an "attractor", it can constrain selection if maximal function lies outside of the attractor. Landscape-attractor diagrams allow us to evaluate the effectiveness of different selection regimens, including screening communities comprising one random clone from each species.