Fig. 3.

Comparison of fluxes simulated with different measurement errors. Panels depict a flux magnitude, b median error, and c percent non-significance. Simulated data was drawn from the stoichiometric model described in the Methods section, constrained by 99 % confidence intervals on fluxes observed between 18 and 80 hours post inoculation. 100 balanced flux profiles were generated with 100 random generated sets of measurement error applied to each. Each balanced flux profile was ordered according to increasing absolute flux magnitude to generate an associated rank from 1 to 45