Research led by the University of Chicago shows an artificial intelligence (AI) driven “digital twin” modeling the infant microbiome can predict neurodevelopmental problems later in infancy.
Using data on very early gut microbiome composition from fecal samples from preterm babies the digital twin predicted later microbiome composition and associated neurodevelopmental deficits with a good degree of accuracy.
“You can only get so far by looking at snapshots of the microbiome and seeing the different levels of how many bacteria are there, because in a preterm infant, the microbiome is constantly changing and maturing,” said Ishanu Chattopadhyay, assistant professor of medicine at the University of Chicago and lead author of the Science Advances study, in a press statement.
“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.”
The research is still at an early stage, but if verified the team believes it could help predict which babies may need microbiome transplantation at an early stage to help improve their neurodevelopment.
“Increasing evidence suggests that microbial dysbiosis contributes to the development and progression of numerous diseases, ranging from facilitating essential digestive processes to regulating the central nervous system through the microbiota-gut-brain axis,” write the authors.
“While the microbiome’s role in brain development and the significance of microbial dysbiosis in neuroinflammation and neurodevelopmental disorders have been observed, including in preterm births, the specific mechanistic pathways operating along the gut-brain axis are yet to be fully understood.”
To try and improve knowledge of this area, Chattopadhyay and colleagues used 16S ribosomal RNA profiles extracted from 398 fecal samples taken from 88 preterm babies to inform and train the digital twin model. The data came from babies who went on to develop neurodevelopmental problems, as well as babies who developed normally, allowing the AI to learn to predict potential developmental issues in newborn babies.
The team found that the digital twin was able to predict risks for suboptimal development and poor head circumference growth with a 76% area under the receiver operator characteristic curve score. It had a 95% positive predictive value and 98% specificity at 30 weeks gestation.
The researchers calculated that early microbiome transplantation could potentially have helped around 45% of the cohort avoid developmental problems, but this would need to be validated in future work, particularly as incorrect supplementation could have adverse effects.
“You can’t just give probiotics and hope that the developmental risk is going to go down,” Chattopadhyay said. “What you are supplanting is important, and for many subjects, you also have to time it precisely.”
The researchers say the digital twin model has the potential to focus research on a smaller number of potential treatments or treatment targets in the gut microbiome. This could reduce the time for development of therapies significantly compared with current timelines.
