Mantis Biotech is creating “digital twins” of humans to help solve the problem of availability of medical data

Large language models trained on massive datasets can accelerate genomics research, simplify clinical documentation, improve real-time diagnosis, support clinical decision-making, accelerate drug discovery, and even generate synthetic data to enhance trials.

But their promise to transform biomedical research often faces a bottleneck: Beyond the structured data that health care relies on, these models face difficulties in cutting-edge cases such as rare diseases and unusual conditions, where reliable, representative data are scarce.

New York-based Mantis Biotech claims to be developing a solution to fill this gap in data availability. The company’s platform combines disparate sources of data to create synthetic datasets that can be used to build so-called “digital twins” of the human body: physics-based predictive models of anatomy, physiology, and behavior.

The company is offering these digital twins for use in data collection and analysis. These digital twins could be used to study and test new medical procedures, train surgical robots, and simulate and predict medical problems or even behavior patterns. For example, a sports team can predict the likelihood of a particular NFL player suffering an Achilles injury based on his recent performance, training load, diet, and duration of activity, Georgia Witchell, founder and CEO of Mantis, explained to TechCrunch in a recent interview.

To build these twins, the Mantis platform first takes data from a variety of sources such as textbooks, motion capture cameras, biometric sensors, training records, and medical imaging. It then uses an LLM-based system to route, validate, and synthesize various data streams, and runs all that information through a physics engine to create high-resolution views of that data set, which can then be used to train predictive models.

“We’re able to take all these disparate data sources and then turn them into predictive models of how people will perform,” Witchell said. “So anytime you want to predict how a human will perform, that’s a good use case for our technology.”

The physics engine layer is key here, Witchell told TechCrunch, because it helps the platform improve the information available by grounding the synthetic data generated and modeling the physics of anatomy realistically.

“If I asked you to do hand posture estimation for someone who’s missing a finger, it would be very difficult, because there are no publicly available datasets of classified hand postures for someone who’s missing a finger,” she said. “We can create this dataset very easily, because we just take our physical model and say, remove finger X, and recreate the model.”

Because the Mantis platform fills gaps in data sources, Witchel believes there is potential for it to be used widely across the biomedical industry, where information about procedures or patients may be difficult to access, disorganized or siled in different sources. She emphasized cutting-edge conditions or rare diseases, where data is difficult to obtain because there are often ethical and regulatory restrictions around including patient data in public datasets, or using it to train AI models.

“You know how when you see a three-year-old running around, and they have a Barbie doll, and they grab it by one leg and slam it on the table? I want people to have that mentality with our digital twins,” she said. “I think this will open people up to the idea that humans can be tested when using virtual humans. I currently feel like people are working with the exact opposite mindset, which makes perfect sense, because people’s privacy has to be respected. In fact, I don’t really think people’s data should be exploited at all, especially when you have these digital twins.”

Currently, Mantis has seen success in professional sports, perhaps due to the need to provide a role model for high-performance athletes. Witchel said one of the startup’s main clients is an NBA team.

“We created these digital representations of athletes, where it basically shows here how this athlete jumped, not just today, but for every day in the past year, and here how their jumps change over time compared to the amount of sleep they have, or compared to the number of times they raise their arms above their head,” she explained.

The startup recently raised $7.4 million in seed funding led by Decibel VC, with participation from Y Combinator, a few angel investors and Liquid 2. The funding will be used for recruiting, advertising, marketing and go-to-market functions.

Witchell said the next step for Mantis is to continue building the technology and eventually launch the platform to the general public, which targets preventive health care. The company also serves the needs of pharmaceutical laboratories and researchers working on FDA trials, with the goal of providing insights into how patients respond to treatments.