Tracking food intake can be a surprisingly difficult task. Dietary questionnaires and food diaries as a part of a clinical study challenge participants to accurately remember and honestly report what they ate throughout the day or week. Plus, these methods sometimes miss fine-grained information about the ingredients within processed foods.
“We thought, to a first approximation, we eat things that contain DNA — mostly plants, fungi, and animals — and that some amount of that DNA probably survives passage through the gut,” said Sean Gibbons, a microbiome researcher at the Institute for Systems Biology. So, he thought, why not measure that DNA directly to find out what people tracking their food were actually eating.
We thought, to a first approximation, we eat things that contain DNA — mostly plants, fungi, and animals — and that some amount of that DNA probably survives passage through the gut.
- Sean Gibbons, Institute for Systems Biology
Gibbons and his team developed a method called Metagenomic Estimation of Dietary Intake (MEDI) that can determine what a person has eaten by analyzing food-derived DNA in stool samples, according to their new study published in Nature Metabolism (1). This approach showed particular promise for tracking protein, carbohydrate, potassium, cholesterol, and vitamin B12 intake.
By using sequenced genomes from hundreds of foods, the team mapped the DNA in stool samples to specific foods and their nutrient content. But, the vast majority of DNA in stool — 99.9 percent — comes from gut bacteria.
"It is a needle in a haystack," Gibbons noted. "It's like 0.1 to 0.0001 percent of all the DNA in stool is coming from food."
Despite this challenge, the team demonstrated that MEDI could reliably detect food components in stool samples, even when present in very small amounts. They validated the method in controlled feeding studies, showing agreement between the amounts of specific foods people consumed and the DNA detected in stool samples. For example, the amounts of protein, fat, cholesterol, and carbohydrate that MEDI detected were significantly correlated with intake. However, the dietary fiber and fat estimates were not accurate.
“MEDI is a promising way to further validate traditional methods and also gain specific additional insights, including the possibility of identifying some specific foods at increasing resolution and also to assess the meals connected with the specific fecal sample under analysis,” said Nicola Segata, a computational microbiome researcher at the University of Trento who was not involved in the study. “We are currently using MEDI in our lab for some of our microbiome investigations, and we expect it to be very useful and open new avenues in our lines of research.”
To show MEDI’s potential in a clinical population, the team identified dietary patterns associated with metabolic syndrome, which is a group of conditions that increases the risk of heart disease, stroke, and diabetes. Their analysis showed that people with metabolic syndrome had high meat and low fruit and vegetable intake, which aligned with previous work (2).
While promising, the team and other experts in the field noted that MEDI has limitations. Steven Heymsfield, a physician-scientist at Louisiana State University who was not involved in the study pointed out practical concerns: "I question whether MEDI has real clinical value, particularly at the individual level.” He noted the additional cost and time MEDI would require compared to questionnaires.
The team hopes to improve MEDI’s accuracy, so that clinicians could at least use it on patients who give stool samples for other purposes. Gibbons hopes to perform additional studies where his team can rigorously control intake before analyzing stool samples so that they understand exactly how well MEDI estimates the nutrient content before digestion.
The most comprehensive way to test MEDI, Gibbons said, “is if we could take a meal that someone's eating, blend it up in a blender, and then do a DNA extraction from that before it goes in, and then feed that smoothie, and then do the extraction on the other side," Gibbons said.
As researchers sequence more food genomes and add them to the MEDI database, the method's accuracy will continue to improve. Gibbons and his team believe that MEDI will become an increasingly valuable tool for understanding the complex relationships between diet, the gut microbiome, and human health.
"We believe that MEDI will be a useful addition to the toolkit of dietary assessment methods," said Gibbons, "especially in studies where stool samples are already being collected for microbiome analysis."
References
- Diener, C. et al. Metagenomic estimation of dietary intake from human stool. Nat Metab 7, 617–630 (2025).
- Thomas, M.S., Calle, M. & Fernandez, M.L. Healthy plant-based diets improve dyslipidemias, insulin resistance, and inflammation in metabolic syndrome. A narrative review. Adv Nutr 14, 44–54 (2023).
