5 minute read

How microbiome analysis could transform food safety

New research shows the microbiome can indicate when there is a potential issue in the supply chain and help predict outbreaks and transform food safety.

New research shows the microbiome can indicate when there is a potential issue in the supply chain and help predict outbreaks and transform food safety.

“Tell me what you eat and I will tell you what you are.”

It was back in 1826 that Jean Anthelme Brillat-Savarin, French lawyer and politician, used this phrase in his book The Physiology of Taste. Fast forward to today, and ‘we are what we eat’ is still used when it comes to describing our microbiome — the community of our gut bacteria, influenced by our every meal. When the food we eat is unsafe, these bacteria wreck havoc inside of us.

But everything we eat also has a microbiome of its own.

It’s these food microbiomes that we wanted to study, using DNA and RNA sequencing to profile microbiomes in the supply chain anywhere along the process from farm to table — in a bid to help improve the safety of the global food supply chain. A complex multi-party network, its monitoring and regular testing can reveal fluctuations indicating an ingredient’s bad quality or a potential hazard. Catching these anomalies at any stage in the supply chain is important, whether in raw ingredients or later in the chain.

To evaluate the use of the microbiome as a hazard indicator for raw food ingredient safety and quality, we used a new type of untargeted sampling. Our team, composed of scientists from IBM Research, the Mars Global Food Safety Center , Bio-Rad Laboratories and consulting professor Dr. Bart Weimer of the UC Davis School of Veterinary Medicine, wanted to see whether the microbiome could indicate a potential issue or deviation from normal in the supply chain and help predict outbreaks.

The latest results, published in the Nature partner journal Science of Food are promising.1

We developed a new technique with specific quality control processes such as bioinformatic host filtering for food and other microbiome studies with mixed or unknown host material. We’ve shown that the food microbiome could indeed help improve food quality control, as well as support a positive shift in microbial risk management, moving from a reactive approach to a predictive and preventative one. This way, it could help ensure safe food by preventing contamination and illness and reducing food waste. Our results have greatly increased the microbial identification accuracy in validation studies on simulated data, and the method is robust for multiple microbiome types, including biomedical, environmental and agricultural samples.

Sequencing our food

As the cost of high-throughput sequencing continues to drop, it has become an increasingly accurate and effective technology to investigate food quality and safety in the supply chain. Many methods employed at scale today involve targeted molecular testing, for example, the Polymerase Chain Reaction (PCR) or bacterial growth assays, or pathogen-isolate genome sequencing.

However, the microbes of interest do not exist in isolation. They are part of an ecosystem of microbes within the microbiome. Microbiome studies have expanded greatly in recent years and are continuing to find new links to human health, the environment, and agriculture.

To kick-off the study, in 2014 IBM Research and Mars co-founded the Consortium for Sequencing the Food Supply Chain . This partnership is supporting best practices in quality and food safety using genomics and big data to better understand the microbiome and its links to global food supply chains.

First, we developed a way to sequence food for accurate authentication of ingredients and detection of contaminants.2 We then expanded that work to characterize the microbiome.1 We sequenced RNA from 31 raw poultry meal ingredients (high protein powder) samples, ranging from a raw materials supplier to the pet food industry, over multiple seasons. This allowed us to identify the baseline of core microbes expected to be present for this sample type. During this monitoring, we also observed that a small number of samples contained unexpected contents. These outlier samples showed marked differences in their microbiomes with additional organisms present, as well as differences in microbial abundance composition.

The microbiome served us as a sort of a detective lens for food quality and safety with greater sensitivity than current tests. Separating the microbial sequences from a highly abundant food sample — be it corn, chicken, or something else — with sufficiently representative reference databases is critical for interpreting the data.

We addressed these challenges using bioinformatic filtering of the food-derived sequences without requiring any a priori knowledge of expected content and by augmenting publicly available microbial reference genome databases.

A functional genomics layer

Beyond the genotypic classification, this work also used the IBM Functional Genomics Platform to layer on biological function data to describe key growth genes involved in Salmonella replication.3 This genetic analysis was used to compare to current culture-based assays and build bridges in our understanding between these existing and emerging technologies.

When the sequenced reads were examined in the context of an augmented reference collection of Salmonella genomes from the IBM Functional Genomics Platform , we observed improved separation between culture-positive and negative samples, demonstrating the utility of developing comprehensive reference databases of microbial genomes.3


  1. Beck, K.L., Haiminen, N., Chambliss, D. et al. Monitoring the microbiome for food safety and quality using deep shotgun sequencing. npj Sci Food 5, 3 (2021). 2

  2. Haiminen, N., Edlund, S., Chambliss, D. et al. Food authentication from shotgun sequencing reads with an application on high protein powders. npj Sci Food 3, 24 (2019).

  3. Seabolt, E. E. et al. IBM Functional Genomics Platform, A Cloud-Based Platform for Studying Microbial Life at Scale. IEEE/ACM Trans. Comput. Biol. and Bioinf. 1–1 (2021) 2