“TrueFood” research database ranks processed foods to unlock healthier choices

US researchers say there needs to be a better understanding and regulation of foods offered at grocery stores. The team developed a publicly available database on the availability of processed foods at three large retailers to inform and empower consumers and policymakers.

The GroceryDB database ranks 50,000 processed foods from Target, Whole Foods and Walmart stores based on ingredient lists, nutrition facts and prices. The results are available on the TrueFood website, where consumers can compare foods. For each food, the database includes a processing score, nutrition facts and an ingredient tree that shows its makeup.

The researchers from Mass General Brigham, US, used the FPro algorithm they developed in 2023 to determine the degree of processing for available foods through machine learning. The higher the resulting processing score, the more ultra-processed a food. 

Co-author Giulia Menichetti, Ph.D., an investigator in the Channing Division of Network Medicine at Brigham and Women’s Hospital, tells Nutrition Insight there is a “lack of large-scale, standardized, open-access repositories in nutrition — resources that have already transformed fields like genomics and transcriptomics, driving precision medicine forward.” 

“Much nutrition research still relies on manual curation, but our study highlights how AI and data science can scale these efforts up. GroceryDB serves as a proof of concept, showing how much we can achieve when data is accessible and algorithm-ready. And this is just the beginning — we’ve barely scratched the surface.”

Menichetti says the database was created to demonstrate how machine learning can unlock insights from real-world, large-scale food composition data. “Our work shows how this information can be translated into actionable metrics, like the degree of food processing, to empower consumer decision-making and inform public health initiatives that strive to improve the overall quality of food environments.”

UPF health impact

Research links high ultra-processed food (UPF) consumption to a myriad of diseases and health risks, such as heart disease, diabetes and obesity. Moreover, the availability of UPFs continues to increase, with studies indicating these foods account for over half of US adults’ calorie consumption. 

At the same time, scientists dispute the validity of processing levels to determine a food’s healthiness, recommending the consideration of a food’s nutritional content instead. 

“Given the abundance of what is currently classified as UPF in our food supply, completely eliminating this category from our daily lives seems unlikely,” says Menichetti. “Additionally, large population studies have begun to show that what is currently categorized as UPF may not be equally bad for our health.” 

In the TrueFood database, consumers can check a product’s processing score, nutrition facts and an ingredient tree that shows its makeup.The FPro processing score captures the complexity of a food’s full nutrient profile and patterns to help consumers better understand how a product is made and processed. The results are published in Nature Food. 

“Unlike other nutrient profiling systems that assess one nutrient at a time or add them up linearly, FPro evaluates the relationships and configurations among nutrients to provide a more holistic assessment. This information is then translated into an intuitive ranking of food items, making it accessible and actionable for consumers.” 

The granularity of this score enables a deeper understanding of consumers’ choices, says Menichetti. “While the sheer number of brands might create an illusion of variety, the true measure of choice lies in the diversity of nutritional properties and the range of FPro scores within each category.” 

She adds that the variability depends on a food category and the grocery store. “For example, Whole Foods offers a broader range of cookies in terms of FPro diversity compared to Walmart and Target, whereas all three stores perform similarly in the pizza category.”

How FPro works

To assess a food’s processing level, Menichetti explains that FPro partially builds on classifications such as NOVA, which is commonly used in epidemiological studies to classify foods ranging from unprocessed/minimally processed (NOVA 1) to ultra-processed (NOVA 4). 

Menichetti explains the algorithm also relies on nutrient facts, ingredient lists and composition data, as nutrient lists are consistently regulated and reported worldwide. Moreover, quantities in unprocessed foods are constrained by physiological ranges and food processing systematically and reproducibly alters concentrations through changes detectable by machine learning. 

FPro provides a continuous score ranging from 0 (representing unprocessed foods like fresh fruits and vegetables and other minimally processed foods) to 1 (representing UPFs, such as instant soups and shelf-stable breads). 

Menichetti says it is unlikely to eliminate UPFs from people’s diet given their abundance and not all are equally bad for health.“Unlike traditional nutrient-focused assessments, FPro does not evaluate individual nutrients in isolation,” continues Menichetti. “Instead, it considers the overall configuration of nutrient concentrations within a fixed quantity of food (100 grams). This approach accounts for the correlations among nutrients and their patterns in unprocessed versus ultra-processed foods.” 

“For instance, while fortified foods may mimic unprocessed foods’ vitamin and mineral content, the algorithm can detect unique nutrient concentration signatures indicative of industrial processing, resulting in a higher FPro score.”

Improving public health

Menichetti highlights that FPro facilitates targeted interventions by enabling nuanced ranking within specific food categories. “These interventions allow for incremental dietary shifts, nudging consumers toward less processed options over time. Preliminary calculations suggest that even small changes in dietary trajectories can reduce disease risk, though more foundational research is necessary to deepen our understanding.” 

GroceryDB also highlights the critical importance of open-access, internationally comparable data in advancing global nutrition security.

“With a relatively small team and a lot of passion, we have created a resource demonstrating the immense potential of data-driven insights in this field. We hope GroceryDB will inspire more data scientists to join the effort and serve as a proof of concept for the transformative understanding that becomes possible when robust data is available.”

She adds that the initiative also targets agencies, governments and grocery stores. “For example, the data can be used to identify cost barriers and guide targeted interventions in specific food groups that are staples for different segments of the population or to enable consumers to discover more affordable options within the same category with comparable FPro scores.”  

Menichetti says grocery stores can use GroceryDB’s insights to review their suppliers and product offerings. The database can enable them to “optimize layouts, shelves and discount policies to better meet the needs of their customers while promoting healthier choices.”