Network size

Smallscale (typically central metabolism)
Difficult to determine network model boundaries
Experimentally and computationally hard to extend for larger networks

Genomescale
Enables finding activity of noncanonical metabolic pathways
Potential false prediction of noncanonical metabolic activities due to the inclusion of reactions with weak biochemical evidence in the network model

Typical experimental inputs

Biomass composition, growth rate, and metabolite uptake and secretion rates

Computational requirements

Isotope tracing measurements; potentially absolute metabolite concentrations

A variety of ‘omics’ datasets
Requires simplifying assumptions for integrative analysis

Mostly hard nonconvex optimization problems solved heuristically

Mostly computationally tractable optimizations (linear or quadratic programming)

Determining a unique flux solution

Typically possible
Assessing uncertainty with confidence intervals

Requires simplifying optimizations (e.g., maximal growth rate)

Compartmentalization

Partially addressed with specific tracers, compartmentspecific markers, cell fractionation

Addressed via simplifying optimization assumptions

Applicability

Inferring fluxes in a specific condition

–

Predict flux adaptation following chemical/genetic alterations
