As drug developers and regulators alike increasingly warm to creative ideas for running clinical trials, a machine learning platform created by three physicists is drawing in $50 million from tech VCs.
Unlearn bills itself as the only company that can generate “digital twins” of patients for use in clinical trials, so that biopharma companies can test their drugs with fewer real patients than they would need to in traditional trials, but get similarly, if not even more, reliable results.
“Our product is not an AI model — it’s a clinical trial,” CEO Charles Fisher wrote in an email interview with TechCrunch.
He divulged that Germany’s Merck KGaA is among three drugmakers using Unlearn’s platform in the design of its clinical trials — although it’s not directly for the digital twin service but prognostic information from the digital twins.
Whereas there have been efforts and guidance from the FDA to use real-world data to support regulatory decisions, none have quite gone as far as incorporating constructed patient profiles.
Unlearn believes its technology can accomplish that through a combination of machine learning and new statistical methods. By taking historical control data from previous clinical trials and training a machine learning model on them, it promises to compute a digital twin for every patient enrolled in a trial so that more patients can be in the treatment arm and fewer in the control arm.
Fisher met his co-founders, Aaron Smith and Jon Walsh, while working at a now-defunct startup that developed motion sensors, TechCrunch reported. They then banded together to see if they could turn clinical trial data sets to build disease-specific machine learning models.
“[Our] intention wasn’t to speed clinical trials — it was pure research into machine learning. But [I] had a pharmaceutical industry background and it soon became apparent that there had been no investment in machine learning as a technology for pharma development,” said Fisher, who was previously a principal scientist at Pfizer. “[Unlearn] evolved through interaction[s] with the pharmaceutical industry.”
The digital twins would incorporate data from trials over time and cover a wide range of biological information, the company added.
“By reducing the size of the control arm, more patients in a TwinRCT have access to a potentially beneficial experimental treatment instead of a placebo,” Dylan Morris, managing director at Insight Partners, said in a prepared statement. “Trials can be run faster and with the same resources so that patients can receive access to more effective treatments sooner.”