Key points:
- An AI-powered “digital twin” of the infant microbiome predicted changing microbial dynamics across development.
- Using fecal samples, the model predicted which babies were at risk for cognitive deficits with 76% accuracy.
- The model developed in this study can be utilized in other systems where large numbers of variables interact with each other such as virus evolution or social phenomena like public opinions.
In infants, imbalances in the microbial community are associated with gastrointestinal diseases and neurodevelopmental deficits. A new study, published in Science Advances, contributes to understanding how gut bacteria interact by developing a novel generative artificial intelligence (AI) tool that models the infant microbiome.
The AI powered “digital twin” of the microbiome predicts the changing dynamics of microbial species in the gut across infant development. Typical wet lab experiments that test for these interactions are time consuming, while AI dramatically reduces the timescale.
“You can only get so far by looking at snapshots of the microbiome and seeing the different levels of how many bacteria are there,” said senior author Ishanun Chattopadhyay, professor at University of Chicago. “In a preterm infant, the microbiome is constantly changing and maturing. So, we developed a new approach using generative AI to build a digital twin of the system that models the interactions of the bacteria as they change.”
Researchers trained a model called Q-net using infant fecal sample data from UChicago’s Comer Children’s Hospital. They then validated its predictions about microbiome development using sample data from Beth Israel Deaconess Medical Center. The model used head circumference measurements to predict babies’ risk of cognitive deficits with 76% accuracy.
Their model also suggested that intervention, including restoring the abundance of a particular bacterial species, could reduce the developmental risk of about 45% of the babies. However, interventions should be carefully considered as the model showed that incorrect interventions can make the risk worse.
Q-net models large numbers of variables that interact with each other, meaning it can be utilized in other systems such as virus evolution or social phenomena like public opinions.
“If you have a large amount of data, you can train this system well and it will figure out what the connections are,” explained Chattopadhyay. “It can capture very subtle differences, so it has a really large number of applications.”