Parameterizing Genome-Scale Models Using Proteomic Data Enables Prediction of Gene Expression

The physiological state of a cell is largely determined by the activity of all the members of the proteome. Here, we develop and apply a workflow to synchronized multiple omic data types with a genome-scale model of metabolism and expression (ME) in Escherichia coli to determine enzyme activities that are fully consistent with the physiological state of the cell. Using omics data sets from various environmental conditions, we found that the protein-to-mRNA ratio (indicative of ribosomal activity) remains condition-independent on an individual gene basis. We also found the turnover rate of hundreds of metabolic enzymes were relatively condition-invariant. These findings represent the largest scale determination of enzyme activities in vivo to date. We show that the determined condition-invariant enzyme activities enable prediction of condition-specific transcriptional activation and repression with much higher accuracy than both the unparametrized ME model and M-modeling methods such as Markov-Chain Monte Carlo sampling.  Therefore, quantitative expression data from one condition enables prediction under many new growth conditions, adding significant value and utility to the data. Thus, ME models provide a multi-scale mechanistic framework that allows synchronization of genomic, transcriptomic, fluxomic, and bibliomic data types leading to prediction of physiological functions.