ABSTRACTFinding drug targets for antimicrobial treatment is a central focus in biomedical research. To discover new drug targets, we developed a method to identify which nutrients are essential for microorganisms. Using 13C labeled metabolites to infer metabolic fluxes is the most informative way to infer metabolic fluxes to date. However, the data can get difficult to acquire in complicated environments, for example, if the pathogen homes in host cells. Although data from gene expression profiling is less informative compared to metabolic tracer derived data, its generation is less laborious, and may still provide the relevant information. Besides this, metabolic fluxes have been successfully predicted by flux balance analysis (FBA). We developed an FBA based approach using the stoichiometric knowledge of the metabolic reactions of a cell combining them with expression profiles of the coding genes. We aimed to identify essential drug targets for specific nutritional uptakes of microorganisms. As a case study, we predicted each single carbon source out of a pool of eight different carbon sources for B. subtilis based on gene expression profiles. The models were in good agreement to models basing on 13C metabolic flux data of the same conditions. We could well predict every carbon source. Later, we applied successfully the model to unseen data from a study in which the carbon source was shifted from glucose to malate and vice versa. Technically, we present a new and fast method to reduce thermodynamically infeasible loops, which is a necessary preprocessing step for such model-building algorithms.SIGNIFICANCEIdentifying metabolic fluxes using 13C labeled tracers is the most informative way to gain insight into metabolic fluxes. However, obtaining the data can be laborious and challenging in a complex environment. Though transcriptional data is an indirect mean to estimate the fluxes, it can help to identify this. Here, we developed a new method employing constraint-based modeling to predict metabolic fluxes embedding gene expression profiles in a linear regression model. As a case study, we used the data from Bacillus subtilis grown under different carbon sources. We could well predict the correct carbon source. Additionally, we established a novel and fast method to remove thermodynamically infeasible loops.
Context‐dependency of synthetic minimal promoters in driving gene expression: a case study
