Unveiling Meropenem Resistance and Co-Resistance Patterns in Klebsiella pneumoniae and Acinetobacter baumannii : A Global Genome Analysis Using ML/DL and Association Mining
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
The increasing prevalence of meropenem-resistant gram-negative bacteria has significantly undermined its effectiveness and has increased treatment failure and mortality rates. The global availability of bacterial WGS data with antimicrobial resistance phenotypes enables large-scale genome analysis to explore resistance determinants. This study investigated the meropenem resistance mechanism in multidrug-resistant (MDR) Klebsiella pneumoniae (KP) and Acinetobacter baumannii (AB) isolates using advanced data analytics approaches.
Methods
We analysed 2,411 KP and 375 AB isolates with meropenem-resistant and susceptible phenotypes from the BV-BRC database. AMR genes and mutations were identified from the isolates using the CARD database as a reference. Significant AMR genes and missense mutations, determined through chi-square tests, were subsequently used to train ML and DL models. The best-performing SVM model was used for sequential feature selection to identify key features. Additionally, association mining was conducted separately on the selected features and the antibiotics data.
Results
Notable differences were observed in the proportions of genes contributing to the meropenem resistance mechanism categories between KP and AB, including carbapenemases (4% in KP, 23% in AB), antibiotic efflux (30%, 60%), target alteration (23%, 12%), and reduced permeability (18%, 3%). Mutation frequencies also vary, with antibiotic efflux (26%, 67%), target alteration (64%, 5%), and reduced permeability (7%, 15%). A total of 410 significant features in KP and 211 in AB were identified for model building. SVM-based feature selection pinpointed seven key features in KP and 10 in AB, resulting in 95% accuracy for both. Association mining revealed bla KPC-2 , bla KPC-3 , ble MBL , and aac(6’)-Ib9 as key factors in KP, and bla OXA-23 , Abau_gyrA_FLO|Ser81Leu, and Abau_OprD_IMP|Asn411Asp in AB associated with meropenem resistance. The observed prevalence of AAC genes and the gyrA mutation, along with insights from association mining, reveals the co-resistance of meropenem with aminoglycosides and fluoroquinolones, while oprD mutations imply potential shared resistance across antibiotics.
Conclusion
The analysis of AMR genes and mutations based on resistance mechanisms revealed distinct differences in meropenem resistance between KP and AB. The ML/DL models and association mining approaches identified key resistance features and cross-antibiotic resistance insights. These findings deepen our understanding of meropenem resistance, enabling more precise and effective antimicrobial interventions.
