Neural network-based predictions of antimicrobial resistance phenotypes in multidrug-resistant Acinetobacter baumannii from whole genome sequencing and gene expression

Abstract

Whole genome sequencing (WGS) potentially represents a rapid approach for antimicrobial resistance genotype-to-phenotype prediction. However, the challenge still exists to predict fully minimum inhibitory concentrations (MICs) and antimicrobial susceptibility phenotypes based on WGS data. This study aimed to establish an artificial intelligence-based computational approach in predicting antimicrobial susceptibilities of multidrug-resistant Acinetobacter baumannii from WGS and gene expression data. Antimicrobial susceptibility testing (AST) was performed using the broth microdilution method for 10 antimicrobial agents. In silico multilocus sequence typing (MLST), antimicrobial resistance genes, and phylogeny based on cgSNP and cgMLST strategies were analyzed. High-throughput qPCR was performed to measure the expression level of antimicrobial resistance (AMR) genes. Most isolates exhibited a high level of resistance to most of the tested antimicrobial agents, with the majority belonging to the IC2/CC92 lineage. Phylogenetic analysis revealed undetected transmission events or local outbreaks. The percentage agreements between AMR phenotype and genotype ranged from 70.08% to 89.96%, with the coefficient of agreement (κ) extending from 0.025 and 0.881. The prediction of AST employed by deep neural network models achieved an accuracy of up to 98.64% on the testing data set. Additionally, several linear regression models demonstrated high prediction accuracy, reaching up to 86.15% within an error range of one gradient, indicating a linear relationship between certain gene expressions and the corresponding antimicrobial MICs. In conclusion, neural network-based predictions could be used as a tool for the surveillance of antimicrobial resistance in multidrug-resistant A. baumannii.

Keywords: Acinetobacter baumannii; deep neural network; gene expression; phenotype prediction; whole genome sequencing.

Fig 1

Sankey diagram showing the molecular epidemiological characteristics of 518 A. baumannii in this study. The lines are drawn connecting STs, collection date, K/O typing and distribution of carbapenemase gene. ST_Pas and ST_Oxf represented the Pasteur and Oxford schemes for A. baumannii, respectively.

Fig 2

The minimum spanning tree of 518 A. baumannii isolates based on the cgMLST profiles in this study. Panels (A) and (B) represent the Pasteur and Oxford schemes for A. baumannii, respectively.

Fig 3

Phylogenetic tree of A. baumannii based on core genome SNPs. Panels (A) and (B) represent the phylogenetic relationship of A. baumannii ST208 and ST195 lineage, respectively. The external color bars represent clades based on cgSNP strategy, K/O typing, and isolation year.

Fig 4

Heatmaps based on sensitivity and specificity of AMR genes for predicting phenotypic AST (S/R) of A. baumannii. Panel (A) represents the sensitivity of the target genes to the antimicrobial phenotype and (B) represents the specificity of the target genes to the antimicrobial phenotype in the current study; S/R, susceptible/resistance.

Fig 5

Workflow of the deep neural network in the study. Isolates were divided into training, validation, and testing data sets. The presence and absence of antimicrobial resistance genes and mobile genetic elements were found using the ResFinder and ISsaga.

Fig 6

The loss plots and ROC curves for the models of validation data in the neural network are shown. The plots on the left display loss plots, and the plots on the right show ROC curves. Models with the same letter prefix indicate the same model. (A) AMK: amikacin, (B) CRO: ceftriaxone, (C) GEN: gentamicin, (D) MEM: meropenem, (E) CAZ: ceftazidime, (F) FEP: cefepime, (G) IPM: imipenem.

Fig 7

The duplication of AMR genes and their expression in A. baumannii strains. Green, orange, and pink represent the single-copy, double-copy, and three-copy. Statistical analysis was conducted using the Wilcoxon rank-sum test. The statistical significance of the tests was determined based on the P-value. Symbols "*" and "****" indicated P < 0.05, and 0.0001, respectively.