New data saving model accurately predicts individual glycemic responses

New data saving model accurately predicts individual glycemic responses

When you eat a snack—a meatball, say, or a marshmallow—how does it affect your blood sugar? It's a surprisingly difficult question: The body's glycemic response to different foods varies depending on individual genetics, microbiomes, hormonal fluctuations, and more. For this reason, providing personalized nutritional advice that can manage diabetes, obesity and cardiovascular disease, among others, requires costly and intrusive testing, making it difficult to deliver effective care at scale.

In a paper in theJournal of Diabetes Science and TechnologyResearchers at the Stevens Institute of Technology offer a new approach: a model savings model that can accurately predict individual glycemic responses without the need for blood draws, stool samples or other unpleasant tests. The key to their approach? Track what people actually eat.

It may sound obvious, but so far most research has focused on macronutrients like grams of carbohydrates instead of the specific foods people eat. We have shown that by analyzing food types it is possible to make highly accurate predictions with far less data. “

Dr. Samantha Kleinberg, Color Chair Professor of Computer Science

The team of Dr. Kleinberg examined two data sets that included both detailed food diaries and continuous glucose monitor data for nearly 500 people with diabetes (in both the US and China). Using existing food databases and Chatgpt, they classified each meal by macronutrient content and also used the structure of foods (for example, meats are more similar to cheeses) so that they can distinguish between nutritionally equivalent foods.

By training an algorithm using dietary data and food characteristics, as well as some demographic details, the team was able to predict each person's glycemic response to each food with virtually the same levels of accuracy in previous studies that included detailed microbiome data and other calculated information.

“We still don’t knowWhy"Including food characteristics makes a big difference," says Dr. Kleinberg. It's possible that food information is a proxy for micronutrients that drive glycemic responses, or that the physical characteristics of certain foods cause people to eat them differently or digest them differently.

By focusing on food types, the team was also able to examine individual variations in glycemic responses. “Because people eat the same meals over and over again, the data gives us visibility into how individual responses to certain foods change over time,” explains Dr. Kleinberg. The team found that including data on menstrual cycles in their model accounted for much of the intra-subject variation, suggesting that shifting hormone levels could play an important role in mediating individual glycemic responses.

The team's model also accurately predicts glycemic response for both the U.S. and Chinese populations—an important finding because microbiome-based models have often struggled to produce accurate results in different cultural contexts. “We don’t need data on a specific regional population to make predictions,” explains Dr. Kleinberg.

The new model is also powerful enough to predict a person's glycemic responses based on demographic data without the need for tailored training on food logs or other personalized data. As a result, clinicians could potentially nourish the model to provide nutritional advice during an initial meeting with a patient, without the need for laborious food protocols or intrusive testing. “We can make better recommendations if we have more data, but we can get very good results without personal information,” explains Dr. Kleinberg. “This means we can give patients useful advice straight away – and hopefully that will motivate them to keep going.”

Next, the team plans to refine their model with larger data sets and examine whether adding microbiome data increases the accuracy of their model. "That's the big question, because if food information alone gives us everything we need, there may be no need to collect stool samples or do other tests," says Dr. Kleinberg. “This could make personalized nutrition more affordable and accessible to everyone.”

Sources: