Recursion Pharmaceuticals has a prescription for high drug costs: AI and 25 petabytes of data

By Eleanor Laise

Growing interest in AI-powered drug discovery draws some big investors--and regulatory questions

Amid fierce debate about prescription-drug prices and countless proposals to rein them in, there has been a point of widespread agreement: Those prices at least partly reflect the vast amount of time, money and risk involved in developing new medicines.

Salt Lake City-based Recursion Pharmaceuticals Inc. (RXRX) is looking to tackle that problem with its artificial-intelligence models for drug discovery and 25-petabyte biological and chemical dataset. With a warehouse full of robots running millions of experiments per week continuously adding to that trove of data, Recursion is working to build "a foundational model of how biology and chemistry work and interact" that can profoundly change the drug discovery process, Recursion CEO and co-founder Chris Gibson told MarketWatch.

Such a model, Gibson said, can "fundamentally change the price, the cost, the number of medicines in the clinic. And it won't be because it takes less money to synthesize the chemical to put into a pill--it will be because we hope to eliminate the failure rate."

Currently, about nine out of 10 drug candidates that make it into clinical studies fail. And for those drugs that do reach the U.S. market, the journey typically takes 10 to 15 years and costs about $2.5 billion, by some estimates.

With its vision for a more efficient, industrialized drug discovery process, Recursion has lately acquired some powerful allies. The company in July announced a $50 million investment from chip maker Nvidia Corp. (NVDA) and a collaboration with the tech giant to optimize and distribute Recursion's AI models to biotech companies using Nvidia cloud services. Recursion shares, which skyrocketed on the news, have since retreated but are still up more than 5% in the year to date.

AI-powered drug discovery, meanwhile, is also drawing interest from lawmakers and regulators. Sen. Bill Cassidy, a Louisiana Republican and ranking member of the Senate Health, Education, Labor and Pensions Committee, this week released a white paper on AI that underscored some of its potential risks and benefits in healthcare, including drug discovery. Congress should support the continued growth of AI for drug research and development, the paper said, but the U.S. Food and Drug Administration also needs "world-leading expertise to keep up" amid the increasing pace and complexity of drug development.

Senate Majority Leader Chuck Schumer, a New York Democrat, plans to host a wide-ranging AI forum on Wednesday. While legislating on the topic won't be easy, Schumer said this week, "We cannot behave like ostriches sticking our heads in the sand when it comes to AI."

The FDA has also waded into the conversation, releasing two papers this spring on AI and machine learning in drug development and manufacturing. The advancements could help bring effective, high-quality treatments to patients faster, the agency said, but also raise concerns--such as a potential lack of transparency in algorithms that could allow errors or preexisting biases in data to grow even worse.

More than 100 drug and biologic applications submitted to the FDA in 2021 included AI and machine learning components, according to the agency.

There are plenty of potential obstacles ahead. "Trust in AI is a major barrier," according to a recent report on AI in drug discovery by the Boston Consulting Group and Wellcome Trust, a health research foundation. And the new technology still needs to prove its worth in drug discovery, the report said, noting that "drugs developed through AI approaches are now entering the clinic, which will be a critical test" of whether they actually have a higher probability of clinical success and improve on current standards of care.

Some skeptics of AI-powered drug development have also raised broader concerns about AI and the "hallucinations" of large language models like ChatGPT. "I get why people are scared," Gibson said. But Recursion isn't marketing drugs based on machine-learning predictions, he said. Instead, it's using machine learning and AI to make predictions about which compounds to test, then putting the actual compounds into experiments with real human cells and doing all the same studies that the FDA would expect for any compound, he said.

While a number of competitors are also pursuing AI-powered drug development, Recursion also has a unique advantage, Gibson said, in the supercomputer that it built for $25 million in 2021. Making that investment, he said, "was an unpopular, strange thing to do for a biopharma company at the time." But "we think it's a full-on gold rush" to discover better medicines with lower failure rates, he said, and "we are on top of a massive quantity of gold." Data and graphics processing units, or GPUs, he said, are the "picks and shovels."

Some Recursion partners in the pharmaceutical industry have signed on for the long haul. In late 2021, the company announced a collaboration with Roche Holding AG (ROG.EB) and its Genentech unit to rapidly identify new targets and advance treatments in neuroscience and oncology. Under the agreement, Roche and Genentech may initiate up to 40 programs, the companies said.

Recursion also has five programs in its own pipeline, including three phase two candidates. The company announced this week the completion of an early-stage study assessing its investigational non-antibiotic small molecule designed to treat Clostridioides difficile (C. diff) infections. Based on promising results from the phase 1 study, Recursion said it is looking to advance the treatment to a phase 2 study to launch next year.

If Recursion is successful in its broader mission, "we'll be able to develop many medicines for much lower prices," Gibson said. While he said he's not a fan of the Inflation Reduction Act's Medicare drug-price negotiation provisions, "I do believe we have to bring down the price of medicines, especially in the United States, and I think the fundamental way you do that is by understanding biology such that you don't fail 90% of the time in the clinic," he said.

"If we failed eight out of 10 times in clinical development," he said, "we could price all our medicines at half the price of everybody else and still win."

-Eleanor Laise

This content was created by MarketWatch, which is operated by Dow Jones & Co. MarketWatch is published independently from Dow Jones Newswires and The Wall Street Journal.

(END) Dow Jones Newswires

09-08-23 1238ET

Copyright (c) 2023 Dow Jones & Company, Inc.