Fig. 1: Schematic for PRIME model development and performance assessment.
a Schema for integration of gene regulation and metabolism. The gene regulatory network (GRN) models weighted regulatory influences of TFs on regulated genes (RGenes). A subset of the RGenes are enzyme-coding metabolic genes (Mgenes), whose functions are also modeled through gene-to-protein-to-reaction (GPR) mapping in a stoichiometric matrix representation of the metabolic network (MN). PRIME uses the integrated Gene Regulatory Network of Metabolism (GRNM) and a reaction flux influence estimator (ReFInE) to calculate the \(\gamma\) factor, which quantifies how the differential expression of multiple TFs and their weighted regulatory influences on a regulated metabolic gene (RMGene) manifest in altered flux (aw: minimum flux; bw: maximum flux) through the associated metabolic reaction (RMRxn) in a given environmental condition. b Illustration of condition-specific gene phenotype predictions and performance assessment. The example illustrates how PRIME predicts relative growth consequence of single gene knockouts in TFs (e.g., TF1, TF2, and TF3) and RMGenes (e.g., G1, G3, G6, and G7) in different contexts (e.g., Condition 1, 2, and 3). The vertical line in the barplot depicts a user-defined threshold in growth inhibition, below which a gene is deemed essential. Performance of PRIME is quantified using a receiver operating characteristic (ROC) curve based on accuracy of PRIME-predicted essential and non-essential genes in a given condition to experimentally determined phenotype consequences using transposon mutagenesis coupled with sequencing (TnSeq) in the same condition.
