I have created a convenience kinetics model. Now, I want to integrate the transcriptomics data with my convenience kinetics model for altering/weighing the kinetic parameter values. I have read some publications related to this work but can’t get a satisfactory idea of how to do it properly.


I've got a pancreatic model for the study on type 2 diabetic patients. It has a few compartments and the reaction pathways; the species are metabolites. All the reactions are convenience kinetics based. Enzyme is a constant factor, so its production and regulation is not considered. From my model, I'll get the Vm values for all the reactions and then I'm supposed to get the Vm (Vmax of Michaelis-Menten kinetics) values for the type 2 diabetic patients by somehow integrating the transcriptomics data with it. I was asked to use the fold change values but I couldn't find any relevant publication on this topic (my PI has suggested me these but hasn't worked in this area before).

Any leads or references will be highly appreciated.

You can call convenience kinetics rate laws as approximate rate laws or you might have heard about "modular rate laws", they are more or less the same. We use this approximate rate law approach in cases where we don't have the experimental data for all the true rate law parameters.

  • $\begingroup$ Anyway, I have come across people who have been trying to incorporate transcriptomics data in flux balance models. I'm not sure how much that improved the accuracy of the predictions. There are problems at multiple levels. I don't know much about what are the parameters that can be incorporated in FBA but I am a bit familiar with kinetic models. If you can present a minimal model with the details on how you are planning to write the equations and so on, I may be able to give some suggestions. $\endgroup$
    Jul 4, 2019 at 14:09
  • $\begingroup$ I've added that information in the question now. Can you please tell me about those researchers/labs who have worked on their integration? Maybe provide me with the links to their webpages or the publications that you're referring to. It'd be of great help. Regarding the approach, I myself have no idea and have been stuck here from the past week, and have been asking around people at different labs. So, any kind of lead will be appreciated. @WYSIWYG $\endgroup$ Jul 4, 2019 at 14:29

1 Answer 1


I am assuming that in your model, the reactants and products (species) are metabolites and each reaction denotes conversion of one metabolite to another.

From transcriptomics, you will get the relative expression levels of different genes. When you have two samples from different conditions you can calculate the differential expression.

A model can be as complex as you can make it but we can start with simple assumptions. As you said, enzymes are assumed to be constant with respect to metabolites and are not dynamically changing. I also guess that you are not considering genes other than enzymes to be affecting the metabolic reactions (you might actually want to consider the solute transporters). Also, you are assuming Michaelis-Menten kinetics for all reactions.

In that case your Vmax (i.e. maximum rate of a reaction) would be kcat × E0 where kcat is the turnover number and E0 is the total amount of enzyme.

E0 can be approximated using the transcriptomics data. However, transcriptomics data is relative and not absolute. If you need absolute quantification, you must have at least one reference whose absolute mRNA copy number is known. Another issue is that the ratio of protein and mRNA expression would not be the same for all genes and it is basically the amount of active protein you need to know. So proteomics would be one step closer.

When you are comparing two conditions, you can use the differential expression (fold change values) to adjust the parameters between the two conditions. For e.g. if you know that phosphofructokinase is 2 fold downregulated in diabetes (compared to healthy case) then you can reduce the Vmax of the corresponding reaction by 2 fold in your diabetes model. However, it all would make sense only if the parameters of your "healthy" model are reasonably close to the biological reality.

Moreover, you still need to know kcat. It cannot be obtained from high throughput studies. You may have to either make some guesses or check out papers/databases. Also, your rate is not always Vmax. To estimate the dynamic rate you must know KM too (when the substrate is in excess you can ignore KM). BRENDA may have some information about these constants for different enzymes.

At a very crude level, you can remove the reactions whose corresponding enzymes have zero expression.

These are some articles on integration of transcriptomics data with FBA (not kinetic models):

Machado and Herrgård actually claim that integration of transcriptomics data does not improve their model predictions:

Also, it is observed that for many conditions, the predictions obtained by simple flux balance analysis using growth maximization and parsimony criteria are as good or better than those obtained using methods that incorporate transcriptomic data.

  • 1
    $\begingroup$ Another alternative is to model Transcription data as very sharp hill function with some fitting parameter. This will essentially mimic Boolean Model and have used successfully in generating regulatory networks. Lot of Eric Davidson's reviews have explained this. For example, Bolouri and Davidson 2002 $\endgroup$
    – Dexter
    Jul 9, 2019 at 12:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .