In order to integrate gene expression data into GEMs, we developed relative metabolic differences (RMetD), which allow for the application of relative gene expressions between CONV-R and GF mice rather than the absolute values, and simulated the metabolic differences using the content of the diet (Dataset EV10 ). First, we set the lower bounds of production of HDL and chylomicrons to 20% (arbitrary value) of their maximum production in GF small intestine model. We performed flux variability analysis for all reactions associated with the significantly (Q-value < 0.05) differentially expressed genes in the model, and found the upper and lower bound of these reactions. A reaction could be associated with more than one gene, and these genes may have different expression trends (e.g. one gene is up-regulated, whereas the other genes are down-regulated). In such cases, we assumed that genes associated with these reactions are not significantly changed (Q-value < 0.05).
Next, for reactions associated with the up-regulated genes in CONV-R mice, we set both the upper and lower bounds of the reactions in CONV-R model 20% (arbitrary value) more than the bound of the reactions in GF model whereas 20% less for the reactions associated with the down-regulated genes. By this way, reactions with up-/down-regulated gene expression were able to carry more/less fluxes. By adding these new constraints for each reaction in CONV-R and GF mice together with the use of the content of the diet, we predicted the fluxes of both models by maximizing the production of chylomicrons and minimizing the sum of fluxes. For all arbitrary values, we performed sensitivity analysis by using different values and obtained similar results.
All the simulations were carried out using RAVEN toolbox (Agren et al, 2013 (link)). RMetD source code was implemented in MATLAB and RAVEN toolbox, and it is publically available athttps://sourceforge.net/projects/relative-metabolic-differences/files/MiceStudy/ .
Next, for reactions associated with the up-regulated genes in CONV-R mice, we set both the upper and lower bounds of the reactions in CONV-R model 20% (arbitrary value) more than the bound of the reactions in GF model whereas 20% less for the reactions associated with the down-regulated genes. By this way, reactions with up-/down-regulated gene expression were able to carry more/less fluxes. By adding these new constraints for each reaction in CONV-R and GF mice together with the use of the content of the diet, we predicted the fluxes of both models by maximizing the production of chylomicrons and minimizing the sum of fluxes. For all arbitrary values, we performed sensitivity analysis by using different values and obtained similar results.
All the simulations were carried out using RAVEN toolbox (Agren et al, 2013 (link)). RMetD source code was implemented in MATLAB and RAVEN toolbox, and it is publically available at