Researchers at the Australian National Phenome Centre (ANPC) at Murdoch University, working in collaboration with Imperial College London, have built an artificial intelligence model that can tell the difference between people with healthy and unhealthy diets based on the pattern of metabolites in their urine, that outperforms conventional dietary scoring systems. This research was published in the journal Nature, Food.
Director of the ANPC and Pro-Vice Chancellor for the Health Futures Institute at Murdoch University, Professor Jeremy Nicholson, a pioneer of metabolic research, said while poor diet is a major predictor for disease, humans are notoriously bad at accurately reporting what they eat.
Professor Nicholson and his colleagues released a related research paper (also published in Nature, Food) on how individuals respond differently to the foods they eat.
This new paper by lead authors Professor Nicholson and Professor Elaine Holmes, Premier’s Fellow and Head of Computational Medicine at Murdoch, delivers a new approach for assessing diet at a population level through urinary metabolic phenotyping.
The ANPC and Imperial groups analysed the chemistry urine samples taken from two large population cohorts from the USA and UK and used artificial intelligence models accurately predict healthy and unhealthy diets among the groups, without any prior knowledge of their dietary intake.
The results were compared to five different traditional ways of interpreting dietary recall data and proved to be more stable. The paper also presents an interactive Artificial Intelligence tool “NutriomeXplorer” that can be used to investigate complex diet-health related metabolic interactions in population data obtained from thousands of people.
“Urine analysis is the key as its chemical composition tells us a lot about what is happening in the human body,” Professor Holmes said.
“The approach we’ve developed takes less than 10 minutes to collect individual data and get an incredibly detailed picture of an individual’s diet.
“Analysing the metabolic profile of urine is digital and much more rigorous than assessing reported dietary intakes.”
Professor Nicholson said: “The application of this knowledge may alert health professionals to groups at greater risk of serious disease based on their diets, providing new opportunities to develop individual prevention and intervention strategies as well as informing future health policy on healthy eating.”.
“This deeper knowledge at the molecular level may translate into new understanding of the biological pathways that are associated with moving from a healthy to unhealthy metabolic position and hence identify ‘entry points’ for early intervention to decrease risk for a number of chronic disease conditions such as hypertension or diabetes.”
Professor Gary Frost, head of nutrition research at Imperial said: “We have demonstrated the value of high-end rapid molecular screening tools to uncover deep layers of human metabolism relating to diet at the population level – this is one of the most important paradigm shifts in nutrition research in recent years because it works in the real world.”
Source: Murdoch University