Diet Not Working? Let AI Rearrange Your Plate
September 20, 2022
LSU researchers are using artificial intelligence, or AI, to effectively predict individual responses to different diets, which traditionally have been prescribed using a one-size-fits-all approach. LSU’s Pennington Biomedical Research Center recently partnered with LSU Health New Orleans to leverage new technologies in the fight against the obesity epidemic and health disparities in Louisiana by joining the largest-ever national effort to leverage big data science for precision health.
It can be tricky to establish an ideal diet for any individual since there are so many variables. Beyond differences in genetics, metabolism, physiology, microbes in the gut, and behavior—particular to any one person—there are also socioeconomic, cultural, and environmental factors. The complexity of the combined data can make it challenging for healthcare providers to recommend the best diet for a person who suffers from obesity or diabetes, for example, even if it’s well known that what we eat can trigger either good health or chronic disease.
In Louisiana, more than one in three residents has obesity while every seventh adult has been diagnosed with diabetes. While the cost and consequences of diet-related, or metabolic, disease are rather clear, the solutions are often more complicated than recommendations to “eat less carbs, saturated fats, and processed foods,” the LSU researchers argue. For example, some of us live in neighborhoods known as food deserts where healthy options and access to fresh fruits and vegetables are limited.
“We know what the best proposed diets are at the population level, but not yet on the individual level,” said Eric Ravussin, LSU Boyd Professor at Pennington Biomedical Research Center. “We’re now at a point where we can take advantage of new technology, such as artificial intelligence, to uncover individual differences and make better predictions for how a change in diet can impact a person’s health.”
Ravussin is one of the world’s most-cited researchers and one of two principal investigators on the new National Institutes of Health-funded project, called Nutrition for Precision Health. This AI-powered effort builds on a larger research program called All of Us, which aims to collect biomedical data on 1 million people across the United States—a sizeable chunk of the total population—enrolling 500,000 so far. Pennington Biomedical recently partnered with LSU Health New Orleans as one of six collaborative clinical centers in the nation to do a deep dive into how different people respond to foods and dietary patterns.
“‘This is what your body does in response to food,’ that’s what we’ll be able to tell our patients,” said Dr. Lucio Miele, geneticist, assistant dean for translational science at LSU Health New Orleans, and co-chair of the Community Engagement National Committee of Nutrition for Precision Health. “All doctors say, ‘Eat healthy and exercise,’ because it plays a positive role in almost any condition short of falling out of a window. Outside of trauma, eating well and moving your body will do wonders for everything, but that’s generic advice and, quite frankly, almost no one does. Having quantitative data, through this study, will make all the difference.”
The study will include more than 1,200 Louisiana residents who are or will become part of the All of Us cohort. Among them, 100 will live and eat all of their meals—following one of three different diets—in a controlled environment at Pennington Biomedical during three periods of two weeks each.
“We’ll get very sophisticated measurements,” Ravussin said. “One size truly does not fit all when it comes to diet, but with continuous glucose monitoring combined with samples of blood, saliva, urine, and hair, we’ll get a much better understanding of the relationship between diet and health. It’s amazing what science can do when, for example, we can determine if you regularly drink soda or eat fish from a single strand of hair.”
The detailed data from the ongoing study will be linked with data from the much broader All of Us research program to derive machine learning algorithms, or AI, for personalized nutrition to combat obesity and metabolic disease more effectively in diverse populations. One of the most powerful aspects of AI is the ability to identify patterns, both known and unknown.
“The idea is to use AI and machine learning to identify the factors in people that make them susceptible—or not—to positive or negative changes in their health after eating different diets,” said Leanne Redman, professor, co-principal investigator alongside Ravussin at Pennington Biomedical, and co-chair of the national governance group of the Nutrition for Precision Health consortium. “At first, these factors could be simple things, such as sex, age, and health condition. But in the future, it will likely be based on what we call ‘omics’ and complex molecular and biological factors that could be measured in a single drop of blood.”
“Omics” are disciplines in biology that study molecules that influence the structure, function, and dynamics of organisms. Examples include the study of genes in genomics, proteins in proteomics, products of metabolism in metabolomics, traits in phenomics, and how ribonucleic acid, or RNA, controls gene expression in transcriptomics.
“As this study progresses, we should get a very detailed mapping of the genes that predispose people to metabolic disease in response to food, including polygenic risk scores,” Dr. Miele said. “Just one whole genome is three tera-bytes worth of data, so using AI and machine learning is the most practical way to capture all of the relevant variables. In Louisiana, we’re also very diverse. We have a lot of metabolic disease and a lot of genetic diversity, so if you want to study how all of these factors come together, you have to do it in Louisiana.”
With the rise of bioinformatics and computational biology, researchers have increasingly been using data science to capture and analyze biomedical information. AI makes these disciplines even more powerful as it can help answer questions researchers might not necessarily think to ask—revealing surprising connections, especially when the data sets are as enormous as in All of Us. Nutrition for Precision Health, meanwhile, will focus on a subset of 10,000 people.
“One of the most exciting aspects of the Nutrition for Precision Health study is the ability to study adults of all ages with a range of health conditions and medication use,” Redman said. “By studying how all kinds of people respond to foods and dietary patterns, nutrition and diet can play a more central role in the prevention and treatment of diseases.”
LSU clinicians will soon be traveling throughout the state to enroll people in the Nutrition for Precision Health study. Participants must be representative of the general population for the results of the study to be applicable to everyone, so while some should have a history with metabolic disease, others should not. The researchers hope to recruit residents who are interested in preventing disease as well.
Better predictive tools and more personalized diet recommendations—both goals of the Nutrition for Precision Health study—could shift the emphasis of medical nutrition therapy from intervention to prevention, according to Kathryn Fakier, who is a registered and licensed dietitian at Our Lady of the Lake Regional Medical Center and director of the dietetic internship program at Franciscan Missionaries of Our Lady University in Baton Rouge.
“Almost 100 percent of the patients I see in the hospital have chronic diseases, and it’s to the point where we’re trying to manage acute complications of the disease instead of preventing disease,” Fakier said. “And while we try to use nutrition to treat chronic disease, it sometimes gets pushed to the side.”
Her colleague, Dr. Tiffany Wesley Ardoin, who is the program director of the Geaux Get Healthy clinical program at Our Lady of the Lake and an assistant professor of clinical medicine at LSU Health New Orleans, agrees.
“Most of my patients come from a vulnerable population and are either overweight or obese...Generally, they don’t have the opportunity to see a dietitian until they develop full-on diabetes, so it would be great for us, as medical doctors, to have more tools to help mitigate disease on the front end. Ideally, every patient should have some type of personalized nutrition recommendation, just like every patient doesn’t get prescribed the same blood pressure medicine when they have high blood pressure.”
Dr. Tiffany Wesley Ardoin, Our Lady of the Lake's Geaux Get Healthy program director and LSU Health New Orleans assistant professor of clinical medicine.
As a dietitian, Fakier is quick to point out that personalized nutrition “definitely is not a new concept for us,” while the data from the Nutrition for Precision Health study could take it a big step further.
“When we’re doing a nutrition education and counseling session with a patient, we’re looking at their lab results, their diagnosis and medical history, surgical history,” Fakier said. “We’re looking at their culture and their background, their access to food; what they have in the kitchen to cook with. Do they even have a stove? We look at all of these factors and come up with an individualized meal plan, so personalized nutrition is not a new concept for us. However, the data from the LSU AI study could be a game changer because we could dive deeper based on gut microbiome and genomics. We could tailor our nutrition prescription more so than ever before.”