Community metabolic modeling combines member genome-scale models into a multi-species model to predict community growth, individual growth rates, and metabolite exchange (cross-feeding and competition). Two facts govern whether the predictions mean anything. First, the community inherits every member's errors: a wrong biomass, missing pathway, or energy-generating cycle in one member distorts the whole community, and members must share a namespace or metabolite exchange silently fails. Second, how the shared space is modeled is the central design choice, and the classic mistake is compartment pooling - a single shared metabolite pool that lets any member use any other's internal metabolites, inventing cross-feeding that requires no secretion. Correct tools (MICOM, SteadyCom) keep members compartmentalized and connect them only through a shared extracellular medium. A separate fork is steady-state (SteadyCom/MICOM) versus dynamic/spatial (COMETS/BacArena) modeling.
pip install micom # abundance-weighted steady-state community FBA; needs a QP solver (HiGHS bundled)
pip install smetana # cross-feeding/competition scores (separate CLI; pairs with CarveMe)
# COMETS dynamic/spatial: cometspy + the COMETS engine (github.com/segrelab/comets)
Inputs: curated member models in SBML sharing one namespace, and member abundances (from metagenomics) for weighting. MICOM ships test_taxonomy() for a runnable example.
Tell your AI agent: - "Build a community model from these member SBMLs and predict community and per-member growth" - "Score cross-feeding and competition between these two species with SMETANA" - "Weight the community by my metagenomic abundances and run cooperative tradeoff" - "Should I use MICOM steady-state or COMETS dynamic modeling for my question?" - "Why is my hand-merged community inventing cross-feeding?"
"I have curated models for five gut species and their relative abundances. Build a MICOM community, run cooperative tradeoff, and report the community growth rate and each member's growth."
"Use SMETANA on my two CarveMe models to compute MRO and MIP, and tell me whether these species mainly compete or cross-feed."
"I want to model succession over time in a co-culture with spatial structure - recommend steady-state MICOM versus dynamic COMETS and explain the parameter cost."
"My community predicts a species growing on a metabolite no one secretes - check whether I built a pooled bag model and how to compartmentalize it properly."