UC San Diego

The growing rates of drug resistant pathogenic bacteria have become a discernible public health crisis, especially in a state in which drug approval rates have dramatically declined. A potential solution to this epidemic is the utilization of a high-throughput screening method that employs metabolic models to predictnew possible drug targets based on synthetic lethality. Here we test the ability of STM_V1.0, a model based on the metabolism of Salmonella Typhimurium LT2, to accurately predict synthetic lethal (SL) gene pairs through experimental validation. We also demonstrate a new method to reconcile the model with experimental data in cases where the model predictions lead to false positive results. Results from the generation of the single and double knockouts corresponding to the predicted SL gene pairs confirm that gltB/gdhA--one of the four selected gene pairs selected for experimental validation--was indeed synthetically lethal, whereas sucC/lpdA, glyA/serA, and ppc/mdh were not. Hypotheses made by the models successfully explained one of these false positive results, revealing that the overexpression of the aceBAK operon was sufficient in rescuing growth in the ppc knockout mutant. These results demonstrate the models have the ability to predict synthetic lethality and also to propose testable hypotheses to resolve discrepancies between in silico predictions and experimental data