Improving food safety using WGS and machine learning

The identification techniques (e.g. conventional typing) used today are time consuming and lack the ability to distinguish between the different strains within the same species. Thus, one can ultimately misjudge the risk of growth of pathogenic bacteria. So-called WGS technology (whole genome sequencing) can alleviate these problems, but the lack of genetic insight and access to advanced computer facilities today prevents dairy companies from using the tools.

By: Grith Mortensen

This project aims to incorporate WGS data from 800 clinical isolations and dairy products into a machine learning tool to identify the virulence level and disinfection resistance of L. monocytogenes strains in real time. The genetics-based results are validated experimentally. With the developed machine learning tool, users can eventually upload raw data sequences to the server, which is hosted at DTU via a freely available web-based interface, after which they receive an easy-to-understand result in 15 minutes per. isolate. The project thus improves and accelerates the decision-making process and food safety management of dairy products.