Charles Fisher, Unlearn.AI: “now is the time to adopt AI-based solutions”

The potential for AI implementation in healthcare can barely be measured, as it can already do what humans do, just countless times better and more efficiently.

Although new technology set to enhance healthcare hits the market nearly every week, AI adoption in his field has been relatively slow, especially when compared to other sectors.

Thankfully, to help create a better tomorrow, solutions like TwinRCTs™ are being developed. Today Cybernews team sat down with Charles Fisher, CEO of Unlearn.AI – a startup developing digital twins for clinical trials – to discuss current challenges in healthcare and how AI technology can address them.

Tell us a little bit about your story. How did Unlearn originate?

I started off my career in academia doing research on theoretical and computational biophysics. Eventually, this area of research began to overlap with machine learning. I left academia for a position at Pfizer working on applications of machine learning to drug development, before moving to San Francisco for a machine learning engineer role at a virtual reality startup.

I met Unlearn’s other founders, Aaron Smith and Jon Walsh, after moving to San Francisco. We had all trained in theoretical physics (broadly speaking) and then moved into machine learning in the tech industry.

Unlearn was born from our observation that most machine learning research at the time was being funded by big tech companies and, as a result, tended to focus on applications relevant to internet companies like natural language and image processing. We figured that by focusing our efforts on a different problem with a different type of data – that is, generative models of longitudinal clinical data – we would discover interesting machine learning techniques that others would miss.

That initial idea developed into our TwinRCT™ solution for clinical trials through a couple of years of R&D as well as interactions with biopharma partners and regulatory agencies.

Can you introduce us to your TwinRCTs™ solution? What technology do you use to accelerate clinical trials?

Clinical trials compare the safety and efficacy of experimental treatments to a currently available treatment option (called the control) to determine if there is a meaningful benefit to the experimental treatment.

Patients enrolling in a trial typically have varying prognoses – some will have better outcomes and others will have worse outcomes regardless of which treatment they get. As a result, clinical trials need to enroll large numbers of patients and randomly assign them to receive the experimental treatment or one of the existing treatment options (which could be a placebo) in order to confidently assess the relative merits of the experimental treatment. This process often takes years and costs hundreds of millions of dollars.

A TwinRCT™ uses an AI model trained on historical data to create a digital prognostic twin for each patient as he/she enrolls in the trial. The prognostic twin provides a prediction of the expected outcome for that patient if he/she were to receive the control. We can leverage this information to design randomized controlled trials that use smaller control groups than traditional designs but maintain other important statistical properties.

What are some of the most challenging issues surrounding clinical trials nowadays?

Identifying, recruiting, and enrolling eligible patients is the most challenging issue in clinical trials. Most clinical trials experience substantial delays in enrollment, which can even jeopardize the success of the trial.

How did the recent global events affect your field of work? Were there any new challenges you had to adapt to?

The Covid-19 pandemic had a large impact on the conduct of clinical trials because many patients were hesitant, or unable, to report to clinical sites during a study. As a result, clinical investigators began to adopt telemedicine and other technological solutions.

In your opinion, what types of organizations should be especially concerned about adopting AI solutions?

An organization could be concerned because they will fall behind or miss out if they don’t implement AI-based solutions, or they could be concerned because prematurely implementing an AI-based solution that isn’t good enough could jeopardize their business.

Every organization should be extremely concerned with both, and getting the balance right is probably the greatest strategic challenge AI poses right now.

Why do you think companies often hesitate to try out new and innovative solutions that would enhance their operations?

In most organizations, failure is punished more than success is rewarded. This leads to an environment characterized by the famous economist John Maynard Keynes as “Worldly wisdom teaches that it is better for the reputation to fail conventionally than to succeed unconventionally.”

You need to spin this around in order to drive innovation forward. Reduce perceived risks of new technologies and maximize their chance of success. For example, Unlearn has been proactive in reaching out to regulatory authorities like the EMA and FDA and has gone through a formal process to obtain a draft qualification opinion from EMA. It’s important for AI companies to take on those difficult tasks themselves to build trust and change the calculus around the status quo.

Where do you hope to see AI used more widely in the future?

Everywhere. AI is simultaneously farther along, and farther away than people imagine. True general intelligence is still a long way off, in my opinion, but more narrow applications of AI are going to be ubiquitous within 5 years.

Since healthcare is your main field of focus, how do you think this sector is going to evolve in the upcoming years?

There needs to be a focus on both costs and value in healthcare. For example, the costs of drug development are too high and, as a result, patients have to pay high prices when it’s not always clear they derive substantial benefits.

One way we can help to address these challenges is by reducing the time and cost of generating evidence about the relative benefits of new drugs.

Share with us, what’s next for Unlearn?

Unlearn is growing rapidly so that we can expand the range of indications for which we partner with biopharma clients to provide TwinRCTs™. In addition, we’re redoubling our R&D efforts to invent even more powerful solutions for the clinical trials of tomorrow.

Enter your email address to download paper.

Click the link to begin download.
Oops! Something went wrong while submitting the form.

Enter your email address to watch the webinar.

Click the link to watch webinar.
Oops! Something went wrong while submitting the form.
Webinars

Why can’t we agree on how to define digital twins in healthcare?

White Papers

Summary of the EMA September 2022 Qualification Opinion for PROCOVA™

White Papers

Case study: TwinRCTs™ reduce control arm sizes for rare neurodegenerative diseases

Learn more about how TwinRCTs™ reduce control arm sizes by up to 22% for Phase 3 trials in HD and ALS.
The European Medicines Agency has qualified Unlearn’s AI-powered method for running smaller, faster clinical trials.
Digital twins seem to be everywhere in healthcare now, but no one agrees on a single definition for them.