Nutrition is an area of great interest with lots of investment and a huge amount of promise but where little progress has been made so far. Although a number of companies are already touting personalized nutrition, supplements and vitamins, there’s little evidence that it’s making people healthier in the long-term. But that’s okay because we need to start with less than perfect options and keep on building our collective experience and evidence to find the best solutions.
AI seems well-suited for this as our response to food involves many factors such as our genes, our environment, our microbiome and other factors that we don’t even understand right now. As such, we need to analyze massive amounts of data to identify patterns and relationships to our health. That’s what AI is well-suited for. One key area to focus on is the analysis of our gut microbiome so that we can determine our response to different foods. ML is showing incredible promise, but studies are hard to carry out because we can’t construct blind studies for nutrition.
The first major development here happened a couple of years back when Eran Elinav, Eran Segal and their team at Israel’s Weizmann Institute of Science published a paper titled “Personalized Nutrition for Prediabetes by Prediction of Glycemic Responses”. In their study, which appeared in the journal Cell, they continuously monitored glucose levels in 800 patients, measuring their responses to 46,898 meals and finding a high amount of variability in the way that they responded to identical meals. This suggests that universal dietary recommendations are of limited use. They also used a machine-learning algorithm to factor in dietary habits, anthropometrics, blood parameters, physical activity and gut microbiota, showing that it accurately predicted patients’ postprandial glycemic response to their meals.
They were also able to validate their predictions in an independent 100-person cohort. On top of that, a blinded, randomized and controlled dietary intervention based on their algorithm led to significantly lower postprandial responses and consistent alterations to the configuration of gut microbiota. When put together, all of these results suggest that personalized diets could successfully modify elevated postprandial blood glucose levels and mitigate its consequences on the metabolism.
There are now a host of companies that offer personalized diets based on the analysis of genomes, microbiomes, and other factors. This is an emerging science with a lot of promise and scientific sense behind it, but we don’t yet have enough evidence to know if it’s better than eating a healthy and balanced diet full of fruits and vegetables.