AI predicts effective treatment for rare disease using existing medications

After combing through 4,000 existing medications, an artificial intelligence tool helped uncover one that saved the life of a patient with idiopathic multicentric Castleman's disease (iMCD). This rare disease has an especially poor survival rate and few treatment options. The patient could be the first of many to have their lives saved by an AI prediction system, which could potentially apply to other rare conditions.

Detailed in a new paper published in NEJM, a group led by researchers at the Perelman School of Medicine at the University of Pennsylvania used an AI technique called machine learning to determine that adalimumab-a monoclonal antibody which is FDA-approved to treat conditions ranging from arthritis to Crohn's disease-was the "top-predicted" new treatment that was likely to work for iMCD.

In parallel, experiments performed by the study team also found that the specific protein that adalimumab inhibits, called tumor necrosis factor (TNF), , was potentially playing a key role in iMCD. They detected elevated TNF signaling levels in patients with the most severe forms of iMCD. Further analysis showed that immune cells from iMCD patients produce more TNF when activated than healthy individuals.

Taking these findings together, the study's senior author David Fajgenbaum, MD, an associate professor of Translational Medicine and Human Genetics, and the physician of the patient in the study, Luke Chen, MD, a hematologist at Vancouver General Hospital in Vancouver, BC, decided to try this TNF inhibitor for the first time in an iMCD patient.

"The patient in this study was entering hospice care, but now he is almost two years into remission," said Fajgenbaum, who is also the co-founder of a non-profit called Every Cure. "This is remarkable not just for this patient and iMCD, but for the implications it has for the use of machine learning to find treatments for even more conditions."

The process of using an existing drug for a purpose other than its initial intentis called drug repurposing. Many diseases may appear very different-in symptoms, prognosis, or even cause-but they could share some underlying links in the body-such as common genetic mutations or molecular triggers-and can therefore be treated with the same drug.

Fajgenbaum has iMCD himself and, through his research, found his own life-saving, repurposed treatment more than a decade ago that has kept him in remission since. The experience inspired him to join the faculty at the University of Pennsylvania and to co-found Every Cure to unlock more life-saving repurposed treatments. The organization seeks to harness artificial intelligence's ability to examine astronomical levels of data to analyze already-approved medications that could serve as potential treatments for people with rare diseases.

The AI platform, used in this study, was built upon pioneering work by study co-authors Chunyu Ma, a research assistant, and David Koslicki, an associate professor of Computer Science and Engineering, Biology, and the Huck Institute of the Life Sciences, both of Penn State University.

One more try

The patient described in this study was heading into hospice care because multiple treatments had failed him over time.

Idiopathic multicentric Castleman's disease is a cytokine storm disorder. Cytokine storms are characterized by an exaggerated and harmful response by the immune system in which too many inflammatory cytokines (proteins in the immune system that play a role in the ways cells communicate with each other) are released and can damage the body's tissues and organs. Those with iMCD can, as a result, experience swelling of the lymph nodes, inflammation throughout their body, and life-threatening multi-organ failure.

The study's patient had experienced many of these issues until he was treated with adalimumab.

The next steps

While Castleman's disease is relatively rare-about 5000 are diagnosed in the US each year-the findings of this study could save the lives of many more.

There are probably a few hundred patients in the United States and few thousand patients around the world who, each year, are in the midst of a deadly flare-up like this patient had been experiencing. More research is needed, but I'm hopeful that many of them could benefit from this new treatment." David Fajgenbaum, MD, associate professor of Translational Medicine and Human Genetics

The study highlights the importance and power of combining a range of scientific approaches, versus using AI, laboratory work, or clinical research methods alone.

Moving forward, Fajgenbaum and his team are gearing up to launch a clinical trial this year of the effectiveness of another repurposed drug, this one a JAK1/2 inhibitor, on iMCD.

This study was supported by grants from the National Heart, Lung, and Blood Institute (R01HL141408), the Food and Drug Administration (R01FD007632), from the National Center for Advancing Translational Sciences (NCATS) (OT2TR003428), as well as Advanced Research Projects Agency for Health (agreement no. 140D042490001), the Castleman Disease Collaborative Network, Every Cure, BioAegis, and the NCATS Biomedical Data Translator Initiative; the Chan Zuckerberg Initiative, Lyda Hill Philanthropies, Arnold Ventures, the Elevate Prize Foundation, and the Carolyn Smith Foundation.

Mumau, M. D., et al. (2025). Identifying and Targeting TNF Signaling in Idiopathic Multicentric Castleman’s Disease. New England Journal of Medicine. doi.org/10.1056/nejmc2412494.