Around six weeks ago, University of Pennsylvania medical professor David Fajgenbaum received a call from the family of a man in Washington state who was dying from a rare blood disorder called POEMS syndrome.
“And from our discussion, I learned that the patient was about to be transferred to hospice within the next day or so,” Fajgenbaum said. “And that the doctor really felt that they were out of options.”
Fajgenbaum isn’t an expert on POEMS, but the physician-researcher does run a nonprofit called Every Cure, which focuses on finding new uses for existing drugs. Fajgenbaum proposed three drugs that are often used to treat multiple myeloma, which he knew from Every Cure’s work was similar to POEMS.
In a last-ditch effort, Fajgenbaum said the patient’s doctors decided to give it a shot.
“And they saved his life. He got out of ICU, he was off [vaso]pressors, he’s off life support — he’s out of the hospital right now. And so he is sort of the first person that we’re putting this to test with.”
Fajgenbaum’s talking about an ambitious project that, last week, was announced to have received $48.3 million in federal funding — an AI-powered platform called MATRIX designed to match diseases with existing, FDA-approved drugs.
“The upside is huge,” Fajgenbaum said. “There’s so many people that could benefit from existing drugs — and rapidly too, right? It’s not like it’s a new technology that’s going to help people 20 years from now. These are the drugs that are at our pharmacy that can help people tomorrow and next year.”
A mission born from personal experience
Every Cure only started two years ago, but its seeds were planted far earlier — all the way back in 2010, when Fajgenbaum, a third-year medical student at the University of Pennsylvania, was struck with a rare disorder called Castleman Disease.
“I became critically ill out of nowhere,” Fajgenbaum said. “I was so sick, I had my last rights read to me because my doctors were sure that I wasn’t going to survive.”
Despite multiple rounds of chemotherapy, the flare-ups kept returning. When Fajgenbaum failed to respond to the only drug in development for Castleman Disease, he realized he was out of options — he needed to find a new treatment if he wanted to live.
“I knew I didn’t have a billion dollars and ten years to create a new drug from scratch,” he said. “If I wanted to survive, I would need to find an existing drug that I could repurpose for my disease.”
In between his hospitalizations, Fajgenbaum pored through his medical records and data from experiments he’d done on himself, looking for clues to a cure. Finally, he discovered a communication line in his body called the mTOR pathway that appeared to play a critical role in activating his immune system. From there, he found a drug used to treat organ rejection, sirolimus, that was known to shut down the mTOR pathway.
“It’s been around for decades, but it had never been used before for my disease,” Fajgenbaum said.
It worked — this past January, Fajgenbaum marked ten years in remission. Since then, he’s poured his energy into finding other drugs that could be repurposed. Over the past ten years, he and his colleagues have found 17 of them, mainly for treating cancers and inflammatory diseases.
At a rate of about one discovery per year, it was light years faster than the process of developing all-new drugs. But Fajgenbaum knew they could do more.
“We thought to ourselves, what if we didn’t just focus on this small number of diseases?’ He said. “What if we utilized the power of artificial intelligence to look for matches across all drugs and all diseases?”
Harnessing the power of AI
Over the past few years, Every Cure has developed a blueprint for an AI-powered platform that can do what humans can’t — match thousands of diseases with thousands of drugs.
“There’s this massive gap where there’s all these drugs, 3,000 drugs, that are approved, and all these diseases, 22,000 diseases,” Fajgenbaum said. And many drugs work for many diseases, but there’s so much untapped potential. And the reason it’s not being tapped right now is because there’s no profit motive for the vast majority of drugs.”
He says 80% of approved drugs are already generic, which means there’s little profit to be made from repurposing them.
Pharmaceutical companies wouldn’t be interested, but someone else with deep pockets was — a new federal agency called the Advanced Research Projects Agency for Health (ARPA-H), which aims to advance big biomedical and health research ideas.
“ARPA-H was created to really fill the gaps that aren’t being addressed by the pharmaceutical industry or by nonprofit research efforts,” Fajgenbaum said.
Currently, Every Cure’s matchmaking program is in beta mode. They’ll be using the government funding to improve and build the platform up, so it can include as many data resources as possible about drugs and diseases. An algorithm will then trawl the databases and match diseases with potential treatments.
With more than 20,000 diseases and 3,000 approved drugs, that means more than 60 million possibilities — far more than any human could ever process, but not impossible for artificial intelligence.
“It can actually look across all 60 million possibilities, look across the world’s knowledge and actually give us a single score for every drug against every disease in a way that humans could never do beforehand,” Fajgenbaum said.
Since doctors can prescribe any FDA-approved drug off-label, Fajgenbaum says, candidate drugs can theoretically be deployed as new treatments right away.
“The big question is whether the insurance company will pay for it,” he said. “Now, if it’s an inexpensive drug or the drug makes a lot of sense for that disease, the insurance company is likely to cover it. But if it’s a really expensive drug and there’s no data supporting it, the insurance company is probably going to reject it.”
So, even though existing drugs have already been proven safe, clinical trials will likely be held to generate data to prove their efficacy in treating a new disease.
The three-year contract includes annual milestones that Every Cure will need to hit. Within one year, they’ll need to have built a publicly available data source that integrates a variety of sources. Within two years, they’ll need to generate scores for the publicly available drugs. And by three years, they’ll need to identify the top candidates among every drug against every disease.
Fajgenbaum said they’ll also launch a portal where physicians, researchers, and patients can contribute ideas in about a year. Until then, they can be submitted at everycure.org.
“You know, I’m not supposed to be here,” Fajgenbaum said. “I was supposed to die many years ago from a disease that didn’t have treatments available for it. And I’m only alive today because we figured out that a drug that was made for something else could save my life. And so all I can think about all day long is how many more of these drugs are just sitting out there at your neighborhood CVS, at your neighborhood Walgreens that are just waiting to be matched to patients who are suffering.”
Read more from our partners, WHYY.
