Timothy Zhu, Temple University
Genome-scale metabolic prediction depends on organism-internal stoichiometry structure and external environment constraints. We seek small-cardinality sets of nutrient imports that satisfy prescribed growth requirements across multiple conditions on flux prediction.
We formulate this problem as a mixed-integer linear program with binary exchange-inclusion variables and continuous flux variables. Flux balance often adopts non-unique solutions, that a common set of imports therefore supports different feasible flux states. Perturbations of qualitatively specified reference media identify candidates of alternative media, while sensitivity to uptake capacities provides complementary evidence.
Our result is an interpretable workflow of environmental input-reduction, with separate assessments of growth capability, flux-range reduction and flux-range preservation.