Figure 1.
Number of publications of methods for integration of transcriptomic data into constraint-based metabolic models: publications per year (bars); cumulative sum (lines).
Figure 2.
Classification of the methods regarding how they treat the gene expression levels (discrete vs continuous, absolute vs relative) and their intended functionality regarding flux prediction, model building or both.
Figure 3.
Prediction error for all methods.
Distribution of the normalized prediction error for each method across multiple conditions for the different datasets. Each box plot represents the distribution of the prediction error for all conditions in one dataset. Two scenarios are evaluated: prediction of the complete metabolic phenotype (growth, secretion and intracellular fluxes) from measured uptake rates (a–c); and prediction of the intracellular fluxes from the measured physiology (growth, uptake and secretion rates) (d–f).
Figure 4.
Physiology prediction (Ishii).
Predicted and measured physiology: secretion rates (mmol/gDW/h) and growth rate (h−1), for the D = 0.7 h−1 experimental condition from the Ishii dataset.
Figure 5.
Predicted and measured physiology: acetate secretion rate (mmol/gDW/h) and growth rate (h−1) for the two over expression mutants (NADP oxidase and ATPase) from the Holm dataset.
Figure 6.
Physiology prediction (Rintala).
Predicted and measured physiology: secretion rates (mmol/gDW/h) and growth rate (h−1) for two extreme conditions (full aerobiosis, full anaerobiosis) from the Rintala dataset.