Benchmarking Metagenomic Pipelines for the Detection of Foodborne Pathogens in Simulated Microbial Communities

Foodborne pathogens pose a significant public health threat worldwide, despite modern advances in food safety. While molecular detection of pathogens in complex food matrices has gained attention to support tracking and preventing outbreaks, thorough benchmarking is needed to optimize workflows for specific scenarios. This study evaluated the performance of four metagenomic classification tools: Kraken2, Kraken2/Bracken, MetaPhlAn4, and Centrifuge, for estimating pathogen presence and abundance in simulated microbial communities representing three food products. Specifically, we evaluated workflow performance in predicting varying levels of Campylobacter jejuni, Cronobacter sakazakii, and Listeria monocytogenes in metagenomes of chicken meat, dried food, and milk products. Metagenomes were simulated to include the respective pathogen at defined relative abundance levels (0%-control, 0.01%, 0.1%, 1%, and 30%) within the respective food microbiome. Performance evaluations demonstrated that Kraken2/Bracken achieved the highest classification accuracy, with consistently higher F1-scores across all food metagenomes, whereas Centrifuge exhibited the weakest performance. MetaPhlAn4 also performed well, particularly in predicting C. sakazakii in dried food metagenomes, but was limited in detecting pathogens at the lowest abundance level (0.01%). Overall, Kraken2/Bracken and Kraken2 exhibited the broadest detection range, correctly identifying pathogen sequence reads down to the 0.01% level, whereas MetaPhlAn4 and Centrifuge had higher limits of detection. Our results highlight Kraken2/Bracken as an effective tool for pathogen detection, with MetaPhlAn4 serving as a valuable alternative depending on pathogen prevalence. These findings provide crucial insights for selecting metagenomic tools for applications in food safety and pathogen surveillance applications.