For a couple decades now, those involved with clinical trials have complained that the process takes too long, requires too much money, and all too often the results don’t end up well—leaving the average successful new drug costing well-over $2.6 billion and many years of development. In the meantime, people continue to get sick and die—often far too early in life. But nothing too profound really changes, and hence as a famous scientist once quipped: “the definition of insanity is doing the same thing over and over again, but expecting different results.” What if to follow Mr. Einstein’s advice requires not learning some new approach but actually “unlearning” the existing one? This is probably why a new San Francisco-based startup secured $12 million in Series A financing. Because through the use of “Digital Twins,” the vendor Unlearn.AI (Unlearn) hopes to help the industry transcend what they do today.
TrialSite News briefly summarizes this new firm, their product offerings and market potential under Investor Watch.
Who is Unlearn.AI?
San Francisco-based Unlearn.AI positions itself to be the developer of the first machine-learning platform that creates Digital Twins used to populate Intelligent Control Arms in clinical studies.
Who are the Founders?
The company was founded by two AI-professionals: Charlie Fisher and Aaron Smith. Mr. Fisher, at the helm as CEO, earned his PhD in Biophysics from Harvard University. Fisher combines his academic pedigree (important in life sciences) with pragmatic considerable industry-expertise: undoubtedly he can help progress a sale. He ventured into the world of starts up first as an engineer focusing on machine learning at Lap Motion and as a computational biologist for Pfizer. It was at Lap Motion where he undoubtedly met co-Founder Aaron Smith—who also possess impressive academic credentials with a PhD in Mathematics from University of Pennsylvania. Smith’s last role was Engineering Manager prior to making the move to start the venture.
The Status Quo
The actual time and cash outlay required for a successful drug is subject to debate. But suffice to say that this process represents an expensive and risky business. More fortunes have been lost than one seeking cures to disease. Over a decade is spent in the drug development process. And as it turns out, of course, much of the time in the course of drug development is represented in clinical trials where a drug’s efficacy and safety must be investigated—and these can carry on for years as trial sponsors (or more usually their CROs) must track down and enroll patients to either participate in the study group that receives the investigational drug or the placebo (or standard of care).
Although this process represents a firm underpinning of the scientific random controlled trial, more often than not results in delays and failure. This author spent considerable time with a firm that helped research sites (including the NIH), CROs and Sponsors two decades go. The same patient recruitment and retention problems pervasive back then are still here today—hence the more we do the same thing over and over the more insane the behavior actually becomes. The Unlearn team refer to some studies that point out that 86% of clinical trials fail to complete patient enrollment on time; while 19% of trials fail because they never enroll enough patients.
Back to Mr. Einstein’s premise that unless a trial sponsor makes a clean break from the ways of doing things then well, they could just be insane. But luckily some smart scientists have developed a digital elixir of their own that may just help the industry consider “unlearning” their existing approach to try something novel.
What does their product do?
Using advanced algorithms, Unlearn’s “Intelligent Control Arm” incorporates AI-generated digital subject data into a clinical trial to reduce the number of subjects who receive the placebo. So these entrepreneurs—and now serious investors—are banking on the ability to essentially digitally simulate the placebo group—a somewhat radical departure from the status quo. In this new world, the Intelligent Control Arms are populated with “Digital Twins,” which according to the venture’s website:
· Vastly reduce the time needed for patient recruitment
· Increase study power by reducing variability
· Provide individualized information about response to treatment
These potential benefits are powered by the proprietary “DiGenesis™ “ process, which creates Digital Twins that are statistically indistinguishable from actual subject data. In an incremental advancement from real-world data research today, DiGenesis involves three steps, including 1) curated clinical data, 2) employing sophisticated machine learning, and 3) applying rigorous statistical analysis. Underlying AI concepts are captured in academic papers involving the two co-founders here.
More on the Concept of ‘Digital Twins’
An interesting concept and core underlying premise in this upstart’s value proposition. One can model patients or groups of patients without actually entering patients in the trial. By taking the data from the large groups of patients and running them through simulation using new therapeutics in the data that’s already been collected from the patients as well as data from previous trials, there is no doubt that this represents a disruptive concept to the status quo governing clinical trials protocol design today. But a growing cadre believe it is absolutely necessary.
Founder Fisher has noted although seemingly a stretch the concept is a scientifically sound concept that the industry is ready to embrace. Fisher notes, “A digital twin is an exact match at baseline can be created for each subject in a trial.” Fisher contends “Digital twins enable faster recruitment with higher powered, individualized trials [capable of generating]additional data” from what turns out to be the same or fewer numbers of patients.
This venture will fist focus on neurological diseases, starting with Alzheimer’s Disease and Multiple Sclerosis as there is great need for new treatment options and fundamental challenges finding patients—and, of course, there has been a series of high profile failed trials.
Open for Business
TrialSite News couldn’t’ find extensive evidence of any substantial deals in place (e.g. white papers or case studies), although there was a few articles involving the company. But this is understandable: this noble startup is trying to get a big, conservative industry to actually unlearn what it is doing today. This takes time, education and lengthy sales cycles, including rigorous scrutiny of the idea, the product and the processes—after all Clinical Quality Assurance (and ultimately regulators) must offer the greenlight.
Who are its Investors?
· 8VC (Principal Francisco Gimenez, PhD on Board)
Although $12 million is a nice start in high income locales, such as San Francisco, it goes fast. The company will need to secure serious business with large pharmaceutical sponsors and CROs. Artificial intelligence requires data and accessing substantial amounts of proprietary data isn’t necessarily easy—nor straightforward. Moreover, VC-backed firms quickly find themselves under enormous, undue pressure to grow faster than the actual market can accommodate—that the actual adoption of disruptive technologies or processes—such as we find with this situation—can take years. Hence, the actual business model required for success probably will look differently than what is envisioned by the founders and their investors. The founders will have to help their VC partners unlearn as well.