Medical Digital Twins: a New Frontier

Last year, a new product for the treatment of type-1 diabetes came on the market: a "digital twin" of the human pancreas. The patient is outfitted with a bloodstream sensor and an insulin pump. The sensor continuously sends data about insulin levels to a device that looks a bit like a cellphone and that runs a mathematical model of glucose metabolism. The model is calibrated to the patient's health status and individual characteristics, such as gender, age, weight, and activity level. The model is linked to a closed-loop control algorithm to drive the pump, which when needed injects the required amount of insulin.

Not only does the digital twin free the patient from the need to pinprick for blood samples several times a day, it also optimizes the amount of insulin administered—just like a healthy human pancreas.

With the success of this kind of model, researchers are starting to envision development of a full-blown "medical digital twin," a software instantiation of the total health status of a person.  One leader in this effort is Reinhard Laubenbacher, director of the Laboratory for Systems Medicine at the University of Florida.

The challenges of medical digital twins are enormous, but Laubenbacher, who received his Ph.D. in mathematics from Northwestern University in 1985 and has spent the past 20 years in systems biology, is ready for it.  "As they say, go big or go home," he said.  "At this stage in my career, my life, that's what I need to do."

Digital twins are used extensively in industry. For example, a digital twin of a jet engine draws on real-time data from sensors in the physical engine to make short-term predictions about the engine's functioning. The twin can make adjustments to head off failure or optimize performance, and can identify faulty or failing components to be checked at the next maintenance. The most sophisticated digital twins are able to self-improve, by learning from situations in which their predictions diverge from what actually happens.

A medical digital twin would take health information about an individual, including data from sensors attached to the person's body, and feed that into a model comprising all major biological systems, from the organ to the cellular and even to the molecular level. Doctors could use the digital twin for a variety of purposes, such as predicting how that particular individual might respond to a given treatment.

However, such comprehensive, detailed models are far in the future. As Laubenbacher put it, "We are at step -1."

A more modest yet still ambitious goal is creating a digital twin of the immune system. Last year, Laubenbacher contributed to a paper on this subject that appeared in the Proceedings of the National Academy of Sciences (PNAS). Reflecting the interdisciplinary nature of the problem, Laubenbacher collaborated on the paper with a dozen colleagues from academic departments of both medicine and computer science, as well as a technology company.

When the immune system responds to an invader, it marshals the biological processes of such key players as white blood cells, antibodies, and the lymphatic system. Supporting players include organs like the skin and the lungs. "There are thousands of labs across the world that study [all these] different aspects," Laubenbacher said. These aspects have been modeled using a range of techniques, from differential equations to finite-state machines to artificial neural networks. Ultimately, he said, "You need to put all of these together."

In the traditional approach to systems biology, models of immune-system components would communicate with each other and exchange data. However, the resulting dependencies between the component models causes complications to proliferate in such a way that modeling the overall system becomes unwieldy. When you want to swap out a white blood cell model for a new and better one, for example, and that model has dependencies on 10 others, the program is likely to break down when you try to execute it.

Laubenbacher and his colleagues proposed a more robust approach using a "hub and spoke" infrastructure for an extensible software platform. The spokes are each modeling modules for the components of the immune system. The modules do not communicate with each other, only with a global data space that functions as the hub. Interactions among the modules are not encoded into a fixed model but instead are simulated, for example by using an agent-based model such as that used in the popular platform NetLogo.

"I can take my white blood cell model and just pull it out and put in another one, and the only thing that matters is that the interface between my model and the global data space remains the same," Laubenbacher explained. This approach, he said, allows one "to build a genuinely modular environment that makes it easy to extend and modify." In this way, the immune digital twin can become a true community effort, with researchers all over the world contributing to it and adapting it to their own needs.

Laubenbacher's laboratory has done extensive research into fungal infections of the lung and has developed expertise in modeling the lung's immune functions. The PNAS paper described a new implementation of their model using the hub-and-spoke infrastructure. The result was a vast simplification of the internal complexity of the model, and greater ease in adding new biological components.

The hub and spoke infrastructure seems quite natural. Has it been tried before? Laubenbacher speculated that it might be used in industrial modeling, an area he is not very familiar with, but he is sure it is uncommon in systems biology. "To my knowledge, at least in the space we are working in, this is novel," he said.

James Glazier, a professor of physics, adjunct professor of informatics and biology, and director of the Biocomplexity Institute at Indiana University, agreed that the approach is uncommon in this area. "Historically, computational biology models were developed and refined independently as small cottage-industry efforts by individual research labs, and their components were rarely reused or improved by others," he said. This resulted in a "throw-away mentality" that slowed the translation of research models into medical practice, he said.

"Building the very complex and interconnected models needed for medical digital twins will require computational biologists to work in new ways, developing and validating simulation components in parallel and collaboratively, with each group working on a separate component," said Glazier. He added that Laubenbacher's "simple but elegant proposed software architecture is an extremely promising approach to enabling such shared development, because it allows software modules to be added, subtracted, or replaced without impairing the function of the entire integrated model. Adopting an approach of this kind could greatly accelerate the development of medical applications of digital twin models."

Fulfilling the promise of this approach means coming to grips with massive computational challenges, as well as a lack of data. "We don't have 50 million patients for whom we have a whole bunch of electronic health records and Fitbit data" that one could mine, said Laubenbacher.

Sensitive ethical questions also arise, such as whether privacy laws apply to one's medical digital twin. Laubenbacher's group is bringing discussion of such issues into the early stages of their work.

For the immune system digital twin, an additional hurdle is posed by the current state of research: much remains unknown about the immune system. One might conclude that the time is not right for the creation of an immune system digital twin, but as Laubenbacher points out, the effort to build such a twin will push research forward by pinpointing salient issues.

"I can build this model and do a sensitivity analysis of hundreds of parameters that show up," he said. "And I can tell you that, for this disease, based on our modeling, it is these three things that you really need to pay attention to, and not those 50 other ones…  So even if you have a less-than-perfect or complete model, it can help guide future research."

 

Allyn Jackson is a journalist specializing in science and mathematics, who is based in Germany.