Penn used an AI tool matched a patient’s rare disease with a common inflammation drug. He’s been in remission for over two years.

Doctors were preparing the patient for hospice, having tried every available treatment for his rare lymph node disorder, Castleman disease, which causes extreme swelling that can lead to organ failure.

In a last-ditch effort, a Penn Medicine immunologist ran lab work through an artificial intelligence tool that matches rare diseases with existing medications.

The patient, who was not being treated at Penn, matched with a TNF inhibitor, a drug that blocks the proteins that cause inflammation associated with diseases such as rheumatoid arthritis. Two and a half years later, he is in remission.

The case study suggests that the technology could be the key to unlocking new treatment options for many more people with rare diseases, for which there are few treatment options, said David Fajgenbaum, an associate professor at University of Pennsylvania. He wrote about his findings in a short article published by the New England Journal of Medicine in February.

Fajgenbaum’s nonprofit, Every Cure, has received $108 million in federal contracts and private financing to identify the promising disease-medication matches and make recommendations for further clinical study.

He declined to talk specifics about the four matches Every Cure is eyeing for its first recommendations this spring, but said his team has found surprising potential in old, cheap generic drugs.

In one instance, Every Cure found a match between an inexpensive generic drug and a common form of cancer. Small-scale academic studies had even found the drug could be effective in treating the cancer, but without a financial incentive to market it for cancer treatment, the match has gone largely unnoticed, he said.

“There appears to be a market failure,” he said. “A cheap drug that wasn’t made for cancer but is potentially very promising — this is just not the way our system is designed to work.”

Every Cure uses artificial intelligence to expedite a drug discovery process that is otherwise often left to serendipity. When patients with rare diseases have few treatment options, doctors may — if they have the time and resources — scour medical journals or tap expert networks for leads on other drugs to try.

Every Cure automates the process with an algorithm that reads massive biomedical knowledge graphs — maps of medical information that show how diseases, genes, proteins, medications, and other biological data are connected. The tool looks for bits of data that diseases and medications may have in common that were previously unrecognized.

Each match is rated from 0 to 0.99, with higher scores indicating a match between a disease and drug with more data in common.

» READ MORE: The Penn doctor who wrote the book on fighting his own rare disease

For instance, Every Cure found that a TNF blocker was associated with some of the same genes that are affected by Castleman disease. The match scored an 0.83 rating — comparable to the 0.87 and 0.89 match scores for the known Castleman treatments.

“Humans knew the TNF blocker affected some of these genes, and humans knew Castleman disease involved those genes, but the connection hadn’t been made that those things overlap,” he said. “It required a machine-learning approach.”

Fajgenbaum founded Every Cure in 2022 with his former medical school roommate, Grant W. Mitchell, and research specialist Tracey Sikora, after his own experience with a rare disease.

He was diagnosed with Castleman disease as a medical student at Penn, and experimented in a campus lab with his own blood to try to find an off-label medication that could address his symptoms.

To date, Every Cure’s successes have been among individual patients. The tool matched a patient with POEMS syndrome, a rare blood disorder that causes nerve damage, with a medication commonly used to treat multiple myeloma, a form of blood cancer.

The broader goal is to identify promising matches for particularly hard-to-treat diseases that could be studied more extensively in clinical trials and, ultimately, lead to a new standard of care and treatment guidelines.