(586f) Comparative Structural Analyses of Antimicrobial Resistant K. Pneumoniae metabolic Networks Via Stochastic Block Modeling and Machine Learning

Increased antimicrobial resistance (AMR) is one of the most pressing public health issues of the modern day. Continued overuse of antibiotics and high adaptability of bacterial metabolisms make gaining an understanding of metabolic links to AMR critical in the race to identify novel drug targets and develop a new generation of antimicrobials. While considerable efforts devoted to identifying the genetic underpinnings of AMR are ongoing, few studies have investigated how AMR relates to structural and topological changes in the metabolic network. A stoichiometric network approach could not only illuminate new pathways as drug targets, but, through cross-species comparison of structural characteristics, could also provide a more fundamental understanding of how bacterial metabolic networks reorganize themselves to cope with antimicrobial threats. The prevalence of strain-specific genome scale metabolic models (GEMs) enables this approach, and we have selected 16 strains of Klebsiella pneumoniae, each with varying AMR profiles, for this work.

In order to detect a wide array of possible network structures, Bayesian inference of the stochastic block model (SBM) was employed as a community detection method. Adjacency matrices were constructed using the stoichiometric matrix of each metabolic network, and both unweighted (Boolean) and weighted versions of each graph were considered. A multilabel classification problem was formulated using AMR profiles as labels and the reaction sets as features. Feature importances were then adapted as edge weights during SBM community detection, thereby emphasizing the most relevant structural differences between each strain’s metabolic networks. Analysis of these differences sheds light on the potentially exploitable relationship between topology and AMR. Continuing work will involve conducting this same analysis on GEMs for over 3,000 strains of resistant and non-resistant E. coli, as well as GEMs of Staph aureus, in order to investigate interspecies motifs related to clinically relevant AMR.