Vietnamese researcher uses AI to 'decode' intestinal bacteria

Researcher Dang Thanh Tung's work (University of Tokyo) helps decode how intestinal bacteria communicate, opening a new direction for personalized medicine.

Illustration: ScienceDaily

According to ScienceDaily , in a groundbreaking study, scientists at the University of Tokyo, Japan have applied advanced artificial intelligence (AI) to decode the complex ecosystem of gut bacteria and the chemical signals between them.

The team developed a new Bayesian neural network called VBayesMM, which detects real biological relationships rather than random correlations. The system outperformed traditional models in studies of obesity, sleep disorders, and cancer.

Gut bacteria play a vital role in human health, influencing digestion, immunity, and even mood. The human body contains around 30-40 trillion human cells, while the gut alone has up to 100 trillion bacterial cells – meaning we carry more bacterial cells than our own cells.

These microorganisms not only participate in digestion, but also produce and transform thousands of small compounds called metabolites – “chemical messengers” that influence metabolism, the immune system and brain function.

"We are only beginning to understand which bacteria produce which metabolites and how these relationships change in different diseases," said researcher Tung Dang (Dang Thanh Tung) of the Tsunoda Laboratory, Department of Biological Sciences, University of Tokyo.

If we can precisely map the interactions between bacteria and chemicals, we can develop personalized treatments – for example, cultivating a particular type of bacteria to produce a substance that has a health benefit, or designing therapies that manipulate those substances to treat disease.”

The problem lies in the sheer scale of the data: thousands of interacting bacterial species and compounds make it extremely difficult to find meaningful patterns.

To solve this, the team used AI with a Bayesian approach to detect the bacterial groups that actually influence each metabolite, and also calculated the confidence level of the predictions – helping to avoid misleading conclusions.

“When tested on real-world data on sleep disorders, obesity, and cancer, our model consistently outperformed existing methods and identified bacterial families that matched known biological processes,” Tung added. “This gives us confidence that the system is detecting real biological relationships and not random statistical patterns.”

The ability to quantify uncertainty helps VBayesMM provide more reliable information to scientists. However, analyzing large microbial datasets is still computationally intensive, although this cost will decrease as processing technology improves. The system performs best when the amount of microbial data is larger than the amount of metabolite data; if the number of microbial data is larger, accuracy will decrease.

In addition, VBayesMM still treats each bacterial species as an independent entity, while in reality they interact complexly with each other.

The team is now looking to expand the model to handle more comprehensive chemical datasets, including compounds from bacteria, the human body, and diet. They also want to incorporate “family trees” of bacterial species to improve predictions and reduce computation time.

“The ultimate goal is to identify specific bacteria that could be targets for treatment or nutritional intervention, thereby moving from basic research to clinical application,” says Tung.

With this new AI tool, scientists are getting closer to harnessing the potential of the gut microbiome to develop personalized medicine, opening the way for more precise and effective healthcare in the future.

Back to topic

VNA

Source: https://tuoitre.vn/nha-nghien-cuu-viet-dung-ai-giai-ma-vi-khuyen-duong-ruot-20251111125341462.htm