Chemogenomic model identifies synergistic drug combinations robust to the pathogen microenvironment
Fig 4
Predicting the impact of metabolic environments on drug interactions in A. baumannii using MAGENTA.
(a) Schematic workflow of the approach for predicting interactions in a new bacterial strain using E. coli drug interaction and chemogenomics data. Genes that are common between E. coli and A. baumannii are overlaid onto the E. coli MAGENTA model. The non-orthologous genes were deleted (i.e. they were set to be zero) and interaction outcomes were predicted using the conserved orthologous genes alone. (b) All pairwise interactions among 6 antibiotics in three media conditions for A. baumannii are shown as heat maps. Blue, white or red boxes correspond to synergistic, additive or antagonistic combinations. (c) Comparison of the interaction scores for each drug combination in three media conditions identified combinations that are sensitive to the environment. For example, ampicillin-tetracycline combination is synergistic, additive and antagonistic in glucose, LB and glycerol environment, respectively. (d) Scatter plot comparison of media specific interaction scores predicted by MAGENTA and experimental measurements demonstrate that MAGENTA can robustly predict antibiotic synergy and antagonism in various environments for a new species using E. coli data (Rank correlation R = 0.57, p = 5x10-5).