Study finds digital twins of human lungs enable personalized testing of new therapies
Researchers at the University of Toronto’s Institute of Medical Science have created highly accurate computer models—called digital twins—of human lungs that can predict how a specific lung will behave, opening the door to personalized testing of new therapies.
A digital twin lung is a virtual replica of a real lung being kept alive outside the body using the Ex Vivo Lung Perfusion (EVLP) system pioneered by University Health Network’s Toronto Lung Transplant Program.
The research, published in Nature Biotechnology, was led by Elly (Xuanzi) Zhou, a PhD candidate in IMS, and her graduate supervisors Shaf Keshavjee, a thoracic surgeon‑scientist at UHN, and Andrew Sage, an assistant professor in the department of surgery at U of T's Temerty Faculty of Medicine.
The EVLP technology—co-created by Keshavjee, director of UHN’s Latner Thoracic Research Laboratories—has transformed not only the availability of donor lungs for transplant, but also how those lungs are studied and assessed. Researchers can now examine lung function in unprecedented detail, free from the body’s systemic effects.
“EVLP presents a really exciting and unique opportunity to collect data and specimens that you couldn't get otherwise,” says Sage, who is also an assistant scientist at the Toronto General Hospital Research Institute. “This study shows how EVLP can be extended into innovative new directions.”
While the lungs were maintained on EVLP, they gathered detailed information on breathing mechanics, blood gases, chemistry, imaging, and molecular biology, creating one of the most comprehensive datasets of human lung behaviour to date.
The lead investigators assembled a multi-disciplinary team of experts from U of T and UHN spanning clinical science, biomedical engineering and data science. Bo Wang, chief AI scientist at UHN, helped to develop key machine learning algorithms that integrated directly with the EVLP platform.
“Using clinical data from close to 1,000 donor lungs, we built a digital twin that mirrors how each organ behaves in the real world,” explains Zhou, who is entering the final year of her PhD program. “This allows us to test interventions virtually and compare predicted outcomes with what happens in actual organs.”
The researchers liken digital twin lungs to the car simulation systems long used by the automotive industry.
“When a car manufacturer changes the size of a piston, they don’t test it by building an entire fleet,” says Sage. “They first simulate the change on a virtual engine to see what performs better, then move forward from there.”
The digital twins predicted lung function with high accuracy across several measurements, including airflow and pressure, oxygen and carbon dioxide exchange, fluid buildup, blood chemistry, X-ray patterns, and gene and protein changes. They found, in many cases, that the virtual lungs outperformed small, conventional control groups used in traditional study designs.
The team then tested how well the digital twin system worked using real-world therapeutic data.
They used a blood clot dissolving drug called alteplase to treat the lungs during EVLP, while the digital twins simulated what would have happened without treatment.
The results revealed a clear benefit that conventional approaches failed to detect—alteplase significantly improved blood pressure in lungs suitable for transplantation but showed no benefit in lungs associated with poor outcomes.
“The concept of digital twins in medicine has gone from just an idea a few years ago to a technology that opens a whole new set of promising possibilities,” says Sage. “We now have the tools to use EVLP and machine learning together to more quickly and precisely identify better therapies.”
The study represents a significant step toward precision medicine at the organ level, where therapeutic interventions can be optimized using virtual copies before being applied in clinical settings.
“What excites me most is how this research could be applied beyond transplantation,” says Zhou. “By using digital twins, we can start testing and tailoring therapies for specific lung diseases, such as pulmonary fibrosis, in ways that weren’t possible before.”
The research was funded by the Canadian Institutes of Health Research and the J.P. Bickell Foundation Medical Research Grant. It was also supported by the AI HUB at University Health Network.
