Machine-learning tech startup Unlearn AI has taken the sci-fi concept of cloning a person and applied it to the world of drug trials.
The San Francisco-based company creates digital replicas of participants in clinical drug trials using artificial intelligence. The result, says Charles Fisher, CEO and co-founder of Unlearn AI, will be more-efficient drug trials that generate better evidence.
Cloning the people who participate in pharmaceutical studies isn’t as creepy as it sounds— although the wide-ranging implications, if it works, could transform the future of drug research. Since the beginning of the year, Unlearn has been testing its technology on clinical trials developing treatments for Alzheimer's disease.
Before a person begins receiving the new treatment under investigation, Unlearn uses a machine-learning model that has ingested reams of historical disease data and applies what it learns to the person’s medical history and personal information, creating what the company calls a “digital twin.”
“Basically, we clone that person, and then we simulate what would happen to them if they were to receive existing treatments or a placebo,” Fisher told me.
This allows researchers to then compare the simulation to the actual observations of how the trial drug is affecting the patient, and, in theory, lead to more robust statistical analysis, even though there are fewer participants.
The digital twins concept has been around for decades, helping engineers build and test high-tech products in a virtual space before actually manufacturing them, but Unlearn AI has applied it in a whole new way. Their unique approach has caught the eye of venture capital firms. On April 20, the 15-employee startup raised a $12 million Series A, led by 8VC, bringing their total funding to $17 million.
“The market for clinical trial improvement is massive,” said Francisco Gimenez of 8VC, an investor in Unlearn. “That's why a lot of companies take this on from various ways.”
But Unlearn's approach is different, Gimenez added. While others experiment with tactics to improve participant recruitment and retention, Unlearn’s digital twin makes those solutions unnecessary.
“To paraphrase Henry Ford, while everyone is looking for a faster horse, I'm going to build a car,” Gimenez said.
Another factor that differentiates Unlearn from others is that the cloning process requires explicit permission from the participants, said Craig Lipset, an adviser to Unlearn. “Unlearn has done a nice job of making sure a patient's permission is front and center,” Lipset said.
The new financing will primarily be used to double the size of the company, hiring a number of people both across technical roles, like software engineers and statisticians and data scientists, as well as people in marketing and business development.
Unlearn sells its product to pharmaceutical companies but they’re not focused on revenue growth so much. Their goal over the next one to two years is to partner with as many pharma companies as they can, so they can demonstrate their approach is scientifically valid. From there, Unlearn can look at going into other disease areas and other parts of the market. Currently they are only focused on Alzheimer's disease.
Fisher sees the long-term implications of his technology impacting how health care is delivered. He can envision a far-off future in which sensors on or in the body provide real-time data to your doctor, who could then run models to learn aspects of your physiology. If you had a disease the models could help reveal how it might progress, and could maybe even create simulations of what your prognosis might be under different treatment options.
Fisher started Unlearn AI in 2017 with co-founders Aaron Smith and Jonathan Walsh with the aim of using AI and health data to improve clinical studies. For them, the greatest bottleneck in the medical research process was with the participants.
Fisher told me that a pharmaceutical company will typically spend about $50,000 per patient to run a trial, and a single trial might take 10 years, have thousands of patients and cost hundreds of millions of dollars.
“It is very, very expensive,” Fisher said. “What our product enables one to do is to run trials with fewer patients and then you can run them much, much faster.”
On top of that, the studies place a burden on the patients, and their families, who risk side effects of new drugs and pressure to participate year after year.
“We'd really like to be able to move things along faster, but also to try to make the whole process better for patients, so that fewer patients have to put themselves at risk,” Fisher said.