A System Identification Based Framework for Genome-Scale Metabolic Model Validation and Refinement

Abstract Due to the scale and complexity of genome-scale metabolic models (GEMs), it has been recognized that one of the major challenges of metabolic network modeling is the evaluation and refinement of GEMs. Currently, besides assessing its size and connectivity, the standard approach for GEM validation is to compare model predictions with experimental data under different conditions. For well-characterized organisms, this point-matching approach works well, because their metabolic network structures have been well-studied and well-defined. However, we show in this work that, for a less studied organism, such an approach did not work well as the combination of multiple model errors resulted in good agreement between the model predictions and experimental data over multiple points. To address it, we propose a system identification (SID) based framework for GEM validation and refinement. The validation and refinement of GEMs for a model yeast Scheffersomyces stipitis is used as the case study to demonstrate the effectiveness of the SID framework.