Improved constraints increase the predictivity and applicability of a linear programming-based dynamic metabolic modeling framework
Background: Current metabolic modeling tools suffer from a variety of limitations, from scalability to simplifying assumptions, that preclude their use in many applications. We recently created a modeling framework, LK-DFBA, that addresses a key gap: capturing metabolite dynamics and regulation while retaining a potentially scalable linear programming structure. Key to this framework's success are the linear kinetics and regulatory constraints imposed on the system. Here, we present improvements to these constraints to improve the predictivity of LK-DFBA models and their applicability to biological systems. Method: Three new constraint approaches were created to better model interactions between metabolites and the reactions they regulate. These new approaches (and the original LK-DFBA approach) were tested on several synthetic and biological systems to determine their performance when using both noiseless and noisy data. To validate our framework, we compared experimental data to metabolite dynamics predicted by LK-DFBA. Results: There was no single optimal type of constraints across all synthetic and biological systems; rather, any one of the four approaches could perform best for a given system. The optimal approach for fitting to wildtype data of a given model was consistently the best approach when predicting new phenotypes for that model. Furthermore, many of LK-DFBA's predictions qualitatively agreed with experimental data. Conclusions: LK-DFBA can be improved by using several kinetics constraint approaches, with the ideal one selected based on wild-type training data. LK-DFBA's ability to predict metabolic trends in experimental data further supports its potential for modeling metabolite dynamics in systems of all sizes.