A modeling framework for predicting resistance to cell wall-inhibiting antibiotics and rational identification of antibiotic combinations in S. aureus

The microbial cell wall is a key target for many antibiotics owing to its conserved essentiality across the bacterial kingdom. Modifications of this cellular component, whether structural, chemical, or organizational, are frequently found in antibiotic-resistant isolates. We develop an ODE (Ordinary Differential Equation) model of the peptidoglycan biosynthesis pathway (PGNet), integrate it with transcriptome data from vancomycin-intermediate S. aureus (VISA), and characterize its cell wall in terms of total peptidoglycan content and degree-of-crosslinking. We show that our model predictions correlate with known cell wall phenotypes for VISA. To understand factors determining such resistance-associated cell wall alterations, we train a random forest regression model on the PGNet solution space, from which we identify ddl as the most influential factor impacting cell wall synthesis, in addition to mraY and murA . Based on these insights, we demonstrate that a combination of vancomycin and d-cycloserine (Ddl-inhibitor), or alternatively, fosfomycin (MurA-inhibitor), significantly retards the emergence of vancomycin-resistance in a clinical MRSA strain.