Scientists have created a new artificial intelligence tool that mimics the infant microbiome. This digital twin can create a virtual model to predict how microbial species in the gut change as the infant grows.
The gut microbiome is crucial for the health and growth of babies. Studies have found that imbalances in the microbial community, known as dysbiosis, are linked to gastrointestinal diseases and problems with brain development. Learning more about how gut bacteria interact and their potential effects on health could lead to important breakthroughs.Some of these problems are difficult and time-consuming to solve through traditional laboratory experiments.
Researchers at the University of Chicago have created a new generative artificial intelligence (AI) tool that models the infant microbiome. This “digital twin” of the infant microbiome generates a virtual model that predicts the changing dynamics of microbial species in the gut and how they change as the infant grows. Using data from fecal samples collected from preterm infants in the neonatal intensive care unit (NICU), researchers used the model, called Q-net, to predict which babies were at risk for cognitive deficits. A new study, published in Science Advances, has found that using generative AI can accurately predict cognitive deficits in preterm infants with a 76% accuracy. According to Ishanu Chattopadhyay, PhD, Assistant Professor of Medicine and senior author of the study, traditional methods of analyzing the microbiome may not be sufficient due to the constant changes and maturation of the microbiome in preterm infants. This led to the development of a new approach using generative AI to create a digital twin of the system that models the interactions of the bacteria as they change. This digital twin concept has the potential to be transformative, just like other forms of AI.logy, which combines computer science, engineering, mathematics, and life sciences, is used to mimic biological systems. When it comes to microbiome dynamics, Chattopadhyay explains that the scale is a critical factor. Traditional wet lab experiments that examine bacteria interactions are time-consuming. For example, testing all the two-way interactions of a typical colony with 1,000 species would take over 1,000 years. This doesn’t even account for the even more complex interactions of three, four, or more species that are common.
The Q-net model significantly reduces the testing time for these interactions, identifying ones that may be of interest for connections.Chattopadhyay and his team trained the model using fecal sample data from infants at UChicago’s Comer Children’s Hospital and then checked its predictions using sample data from Beth Israel Deaconess Medical Center in Boston. The model accurately predicted which babies were at risk for cognitive deficits, as measured by head circumference growth, with a 76% accuracy rate.
The model also suggested that interventions, such as restoring the abundance of a specific bacterial species, could lower the developmental risk for about 45% of the babies. However, the authors warn that the model’s predictions should be interpreted carefully.The study also demonstrated that improper interventions can exacerbate the risk.
“Simply administering probiotics and hoping that the developmental risk will decrease is not enough,” Chattopadhyay explained. “What you are replacing is crucial, and for many individuals, timing is also crucial.”
Q-net has the ability to identify potentially promising combinations of bacteria, significantly narrowing down the search for possible treatment targets. If the gut microbiome is like finding a needle in a haystack, Q-net can help researchers locate the one-inch squares where they can find the needles.
Chattopadhyay’s research colleagues, such as co-aErika Claud, MD, a Professor of Pediatrics and Director of the Center for the Science of Early Trajectories, is conducting research using bioreactors to mimic the live gut microbiome environment. This allows for testing of potential interventions and observing the resulting effects.
The Q-net system, at its core, models numerous interacting variables. Chattopadhyay, the researcher, believes that this system can be applied to other systems beyond the microbiome, such as the evolution of viruses or even social phenomena like public opinions.
According to Chattopadhyay, “If you have a large amount of data, you can train this system well and it will figure out what the connections are.”The software is capable of capturing even the most subtle differences, making it suitable for a wide range of applications.
Â
