The degree to which a food or beverage product is processed is of increasing concern to stakeholders, from nutritionists to consumers to policy makers, concerned about the impact of ultra-processed foods on human and planetary health. I’m here.
However, the most widely used system for evaluating food processing, the NOVA classification system, has come under criticism for allegations that its NOVA 4 ultraprocessed food category is too heterogeneous.
In response, Massachusetts researchers “built” and “extended” NOVA to develop a machine learning algorithm that can accurately predict the degree of processing of any food.
What’s wrong with NOVA?na
With a growing body of research suggesting links between ultra-processed foods (UPFs) and health risks, the degree to which foods are processed is gaining attention.
For the most part, we turn to the NOVA system, which was developed in 2009, to classify the degree of processing. NOVA divides food processing levels into four categories. processed foods; processed foods; and ultra-processed foods.
Although an epidemiological study based on the NOVA 4 (UPF) classification has yielded important findings, researchers at Boston’s Northeastern University and Tufts Medical School say its qualitative nature can lead to ‘inconsistencies’ and ‘ambiguities’. suggests that there is a Additionally, they believe it limits research on the effects of processed foods.
Among the criticisms, researchers note that all risks observed for the NOVA classification are in the NOVA 4 class. This represents a “large and disparate category” of ultra-processed foods that limits the ability of researchers to investigate health effects at various stages. of processing.
Elsewhere, the NOVA system has been similarly criticized. At a media briefing last week hosted by FoodDrinkEurope, Gert Meijer, chairman of the European Technology Platform (ETP) Food for Life and deputy head of corporate regulatory and scientific affairs at Nestlé, said the system “understands the relationships I have raised the issue that Between food intake and health”.
Edith Feskens, professor of global nutrition at Wageningen University in the Netherlands, criticized the NOVA system for not being able to distinguish between different “NOVA 4” products such as sodas and bread.
Machine learning with nutrient inputna
In response, researchers turned to FoodProx, a machine learning classifier trained to predict the degree of processing of any food in a reproducible, portable and scalable way.
FoodProX relies on nutrients as input. This is because lists of nutrients in foods are consistently regulated and reported worldwide, the researchers explained. Their amount in raw foods is limited by a physiological range determined by biochemistry.
Moreover, food processing systematically and reproducibly modifies nutrient concentrations through combinatorial changes detectable by machine learning.
FoodProX allows researchers to define a Continuous Index (FPro) that captures the degree of processing of any food. In addition, it helps researchers quantify an individual’s overall dietary quality and ultimately uncover statistical correlations between the degree of processing and multiple disease phenotypes that characterize an individual’s diet. increase.
Researchers calculated the Individual Food Processing Score (iFPro) for over 20,000 individuals with dietary records in a representative US national sample from 1999 to 2006.
Investigation resultna
Findings suggest that individuals with high food process scores are positively associated with the risk of metabolic syndrome, diabetes, and a family history of heart attack or angina.
Those with higher food processing scores also showed lower blood pressure, trunk fat, obesity, blood insulin, triglyceride levels, and “good” HDL cholesterol.
“Higher consumption of more extensively processed foods correlates with lower levels of vitamins in the bloodstream, such as vitamin B12 and vitamin C. researchers noted.
Researchers believe FoodProX will allow them to “build” and “extend” the current NOVA taxonomy. This includes quantifying the degree of food processing among large, “homogeneously” classified groups of ultra-processed foods.
“Given that our algorithm only requires nutritional information, FPro, information already accessible to consumers through packaging, smartphone apps, web portals, and grocery store and restaurant websites, is based on individual diets. helps monitor whether people are dependent on processed foods.”na
sauce:Nature Communications
“Machine learning prediction of food processing degree”
Published April 21, 2023
DOIs: https://doi.org/10.1038/s41467-023-37457-1
Authors: Giulia Menichetti, Babak Ravandi, Dariush Mozaffarian, Albert-László Barabási
