Screening of new detection targets and virulence factors in Cronobacter using pan-genome and machine learning

Cronobacter species are emerging foodborne pathogens, and species identification is essential because the virulence of different species is diverse, with Cronobacter sakazakii (C. sakazakii) being the most virulent. This study aimed to mine new molecular targets for the detection of Cronobacter strains at both genus and species levels. The genus and species-specific target genes were obtained by using pan-genome analysis. The specificity of selected genes for C. sakazakii (outO, nanK and nanC), C. malonaticus (lpfB_1) and Cronobacter genus (zntB and glpG) was evaluated using PCR amplification. The PCR assays had excellent results, suggesting the potential use of these target genes in food inspection and clinical diagnosis. A machine learning approach was applied to a collection of 1612 Cronobacter strains covering seven species. The predictive performance of input features from two different genomic levels (partial virulence genes and all accessory genes) and two machine learning algorithms (Random Forest and XGBoost) were compared. All models showed high performance and the top important genes contributing to C. sakazakii virulence were ranked. The integration of pan-genome data with machine learning models can be used to monitor Cronobacter spp. and to predict C. sakazakii virulence.