Mantis Biotech is creating ‘digital twins’ of humans to help solve the problem of drug data availability

Large-scale linguistic models trained on multiple datasets can accelerate genomics research, simplify clinical documentation, improve real-time diagnostics, support clinical decisions, accelerate drug discovery, and generate synthetic data to improve trials.

But their promise to revolutionize biomedical research often comes in a bottle: beyond the systematic data healthcare relies on, these models struggle in critical situations like rare diseases and rare conditions, where reliable, representative data are lacking.

New York-based Mantis Biotech says it is developing a solution to fill this data availability gap. The company’s platform combines different data sources to create artificial data sets that can be used to create so-called “digital twins” of the human body: based on physics, predictive models of anatomy, physiology, and behavior.

The company releases these digital twins for use in data aggregation and analysis. These digital twins can be used to study and test new medical procedures, train surgical robots, and simulate and predict medical problems or behavioral patterns. For example, a sports team can predict the likelihood of a certain NFL player developing an Achilles heel injury based on their recent performance, training load, diet, and how long they’ve been working out, Mantis founder and CEO Georgia Witchel explained to TechCrunch in a recent interview.

To create these twins, Mantis’ platform first takes data from various sources such as textbooks, motion capture cameras, biometric sensors, training logs and medical images. Then, it uses an LLM-based system to route, validate, and aggregate various data streams, and run all that information through a physics engine to create a high-fidelity rendering of that dataset, which can then be used to train predictive models.

“We’re able to take all these different data sources and turn them into models that predict how people are going to act. So anytime you want to predict how someone is going to act, that’s a great way to use our technology,” Witchel said.

The physics engine layer is important here, Witchel told TechCrunch, because it helps the platform to improve the available information by supporting the generated artificial data and realistically modeling the physics of the anatomy.

“If I were to ask you to estimate the hand position of a person who is missing a finger, it would be really difficult, because there are no publicly available datasets that have hand positions of a person who is missing a finger. We can generate that dataset really, really easily, because we take our physics model and say, remove finger X, update the model,” he said.

As Mantis’ platform fills gaps in data sources, Witchel thinks there is potential for it to be widely used throughout the medical industry, where information about procedures or patients can be difficult to access, disorganized or locked away from various sources. He emphasized critical conditions or rare diseases, where data is difficult to obtain as there are often ethical and regulatory barriers to integrating patient data into public datasets, or using it to train AI models.

“How do you know when you see a three-year-old running, and she has a Barbie, and she grabs it by one leg and smashes it on the table?” I feel that right now, people are working with a completely different mindset, which is completely logical, because people’s privacy should be respected. In fact, I don’t think people’s data should be used at all, especially when you have these digital twins.”

In the meantime, Mantis has seen success in professional sports, perhaps because of the need to model successful athletes. Witchel said one of the school’s main clients is an NBA team.

“We create these digital images of athletes, where it shows how this athlete jumped, not only today, but every single day in the past year, this is how their jumping changes over time compared to how much they lie down, or compared to how many times they raise their arms above their head,” he explained.

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

The next step for Mantis, says Witchel, is to continue to build the technology, and eventually release the platform to the public, targeting preventive health care. The company also works to support pharmaceutical labs and researchers working on FDA trials, which aim to provide information on how patients respond to treatment.