We are excited to bring Transform 2022 back in-person July 19 and virtually July 20 – 28. Join AI and data leaders for insightful talks and exciting networking opportunities. Register today!
Human gut health is closely correlated with mental well-being, various degenerative diseases and chronic inflammation. One big component of this is the balance of microbes across different digestive tract sections. Researchers have made incredible progress in analyzing microbes that come out the end and taking samples near both openings. But thus far, the relative composition of microbes across different sections in the middle has been a bit of a mystery.
An ambitious collaboration in Europe hopes to change that by building a digital twin of the human gut and its connection to various health outcomes. At the Imec Future Summits Conference in Antwerp, Chris Van Hoof, VP of R&D at Imec and general manager of the OnePlanet research center, discussed the company’s effort to build a digital twin of the human gut. Imec’s team is working to accomplish this by combining new sensors, analytics and machine learning techniques with guidance from nutritionists, doctors, medical researchers, chip designers and data scientists.
Medical researchers are making vast progress in building medical digital twins by exploring humans from the outside. But thus far have very little understanding of what goes on inside the digestive tract.
“You can do gastroscopy and colonoscopy and stool analysis, but no one has charted the entire GI tract,” Van Hoof told VentureBeat.
Imec is a cutting-edge R&D hub facilitating industry collaboration around silicon chips, nanotechnology and health. The current effort involves collaboration with researchers at Wageningen University and Research and Radboud University Medical Center. And Van Hoof said they are soliciting input and guidance from others to help guide research and scale-up initial research.
The largest surface
Medically speaking, the surface of the gut, called the lumen, is outside of the body. Doctors consider it the largest surface of the human body, at about 30 square meters. It acts as a filter and gatekeeper for capturing nutrients and excreting waste. And gut health is correlated with diabetes, metabolic health, weight loss, poor immunity, mental health and dozens of diseases.
There are about 500–800 different populations of microbes in the body. The relative balance of these populations is often correlated with different diseases. Various studies have linked several chronic diseases to microbe populations. And many seemingly unrelated diseases share common microbiome signatures, said Van Hoof.
Researchers do not know why people develop a particular combination or how to tune it. In one recent study, a few of Van Hoof’s colleagues ate a diet rich in a range of plants that might affect their gut health. Everyone on the diet saw a change, but the differences were unique to each individual.
The many pieces of a medical digital twin
Elements of the medical digital twin include a new sensor platform built into a smart pill, a toilet that analyzes stools and urine, an intelligent lunchbox, cameras that track what and how people eat and wearables for correlating these measures with stress and mental well-being.
As for the ingestible, a lot of work went into building a resilient package that was small enough to be swallowed, sufficient for measuring the right things and robust enough to survive the harsh environment adapted to break most substances down. They are now working through the regulatory process and how to start the first human trials at the beginning of next year.
The team wanted to be able to take chemical and biological measurements, record physical characteristics like sound and communicate wirelessly. It also needed highly efficient electronics, since it needed to run for up to a week. They also needed to build a kind of GPS for the stomach to track where it moves in the body.
Today, the knowledge of which factors improve or worsen gut health is only known at the general population level. As a result, doctors tend to focus on population-level advice, such as avoiding fatty or spicy foods, which is not always helpful.
“We hope to create a model of the person to see what causes flare-ups to happen virtually so that the person is not a guinea pig for a treatment,” Van Hoof said, “Rather the digital twin is the guinea pig to try out new interventions to see what would work best for them.”
The team is exploring a range of strategies ranging from individualized digital twins to broader digital twins for groups based on genotype, phenotype, or behavioral or microbiome characteristics. Van Hoof said, “There is hope that we will not have to create billions of models. Rather, we could create a few models and then fine-tune these based on the data we capture to make it more manageable.”
Creating these new models will require a different approach than other domains. For example, a cardiovascular researcher can monitor electrocardiogram data to diagnose many diseases without knowing anything specific about the patient. “But there are no gold standards about what the signal should look like for gut health,” he said.
The effort is also looking at the UI side of the digital twin. Most countries have basic health pyramids characterizing the basic elements of a healthy diet. But getting people to eat more of some things and less of others requires a more behavioral rather than an informational approach.
For example, they have created a smart snack box outfitted with cameras to study people’s snacking behaviors. It turns out that many people, particularly frequent snackers, tend to eat more calories than they record due to unconscious eating patterns; and wearable sensors can track mental and stress changes before someone reaches for a snack.
“We see physiological changes before a person craves some sweets. This is part of the model we hope to link to gut health, where we can capture these signals and hopefully steer people towards healthy alternatives,” Van Hoof said.
Building the right data framework
A digital twin of gut health involves bringing together data from many sources and no one is sure how to do this at scale. Down the road, the hope is that the digital twin could work with sensors, applications and data infrastructure from many parties. But meanwhile, the group is integrating everything into Imec’s OpenPlanet infrastructure, a low-code platform for health and environmental research.
OpenPlanet includes applications for data collection, data connectivity, storage and analysis. It also supports curated algorithms and applications for various use cases. For example, data from wearables can be shared live between doctor and patient or between a digital twin and a health avatar. This can help democratize access to digital twins on top of prebuilt models, data formats and machine learning workflows managed with appropriate security and privacy guardrails.
Van Hoof believes that prior research on the relationship between gut health and other domains has been hindered owing to different ways of gathering, formatting and analyzing data. They avoid these problems by building and integrating all of the pieces in-house. Tools like federated learning could help support a broader range of sensors, data sources and applications down the road.
Creating a digital twin will require more extensive studies and many partners with medical, commercial and healthcare expertise. Van Hoof said they are looking for other parties to join the effort across food and beverage, pharmaceutical and medical technology companies.