ListPred: A predictive ML tool for virulence potential and disinfectant tolerance in Listeria monocytogenes

ABSTRACT Despite current surveillance and sanitation strategies, foodborne pathogens continue to threaten the food industry and public health. Whole genome sequencing (WGS) has reached an unprecedented resolution to analyse and compare pathogenic bacterial isolates. The increased resolution significantly enhances the possibility of tracing transmission routes and contamination sources of foodborne pathogens. In addition, machine learning (ML) on WGS data has shown promising applications for predicting important microbial traits such as virulence, growth potential, and resistance to antimicrobials. Many regulatory agencies have already adapted WGS and ML methods. However, the food industry hasn’t followed a similarly enthusiastic implementation. Some possible reasons for this might be the lack of computational resources and limited expertise to analyse WGS and ML data and interpret the results. Here, we present ListPred, a ML tool to analyse WGS data of Listeria monocytogenes , a very concerning foodborne pathogen. ListPred is able to predict two important bacterial traits, namely virulence potential and disinfectant tolerance, and only requires limited computational resources and practically no bioinformatic expertise, which is essential for a broad application in the food industry. AUTHOR SUMMARY The contamination of food products with microbes such as pathogenic bacteria is a big concern for the food industry. The rapid detection, characterisation and eradication of microbial contaminants are of utmost importance to ensure safe food products. Fortunately, strict food safety regulations and stringent cleaning protocols prevent the transmission of harmful bacteria to humans. Individual bacteria of the same species can have varying abilities to resist cleaning agents or infect a host, meaning that some pathogen isolates might be more dangerous than others. Novel techniques such as genome sequencing and machine learning can help to determine such differences in individual bacteria. Unfortunately, these techniques require a lot of computational power and expertise that is limited in the food industry. This is why we developed an easy-to-use software tool called ListPred that can be used with few computational requirements and little expertise. ListPred helps food companies to answer two essential questions: how dangerous are Listeria monocytogenes pathogens, and how to get rid of them most efficiently?