Published in mSystems, the proof-of-concept study combined shotgun metagenomics data with AI analysis, successfully enabling milk that was treated with antimicrobials mixed into the same tank as clean milk samples. The efficacy of the AI tool was further validated against publicly available, genetically sequenced datasets from bulk milk samples.

Overall, the researchers demonstrated the approach’s ability to detect contaminants and adulterants in untargeted bulk dairy samples. Untargeted methods are important screening methods that can suggest which samples are unsafe and require follow-up investigations.

More specifically, various AI algorithms were tested against 58 bulk tank milk samples to differentiate between clean, baseline samples and samples representing anomalies like milk from an outside farm or milk containing antibiotics. The study characterized raw milk genomes in greater detail than any other project to date, and showed that there is a consistent set of microbes found as stable elements across samples.

Although traditional analysis of microbial sequencing data was not very effective in differentiating between baseline samples and those containing anomalies, the integration of AI facilitated accurate classification and identification of anomalous microbial drivers.

The project used IBM’s open-source AI technology, Automated Explainable AI for Omics, to process the metagenomic data.