A Required Sensitivity Analysis for Predictive Reliability of Genome–Scale Metabolic Networks: Probing Biomass, Nutrient Conditions, and Specialised Metabolism With Penicillium rubens as a Case Study

The biomass objective function is crucial to the predictive capability of genome-scale metabolic network (GSMN) models. Its definition should theoretically reflect the organism-specific macromolecular composition and stoichiometry under defined environmental conditions. In practice, however, reconstruction decisions-whether documented, reasoned, or arbitrary-often lead to model refinements which can potentially introduce ambiguities and may compromise the reliability of simulation outcomes. To mitigate this issue, we propose that systematic sensitivity analysis should be a mandatory step in GSMN validation. This approach quantitatively assesses the reliability of flux predictions by probing a model's responsiveness to perturbations in its core objective function and environmental inputs. We demonstrate this approach using the fungus model Penicillium rubens iPrub22. First, we evaluate the sensitivity of predicted fluxes to variations in the stoichiometric coefficients of the biomass reaction. Then, we examine the model's metabolic behaviour under alternative nutrient conditions. Finally, we assess whether secondary metabolite production, governed by its own regulatory logic, remains robust to changes in the biomass objective function formulated for growth. Together, these analyses measure the degree to which a model's predictions are sensitive to specific reconstruction choices, thereby establishing a standard for evaluating predictive robustness to parameter uncertainties and functional quality in GSMNs.