Fan Zhang, Ph.D., sees artificial intelligence as a pathway to finding an effective way to combat an intractable enemy: rheumatoid arthritis.

Zhang is an assistant professor in the University of Colorado Department of Medicine's Division of Rheumatology and is also affiliated with the Department of Biomedical Informatics on the CU Anschutz Medical Campus. She is furthering her work in harnessing AI to better predict the onset of rheumatoid arthritis (RA) in particular patients, and a new paper documents the latest steps in her work.

The paper is published in the Journal of Clinical Investigation.

Zhang's research focus is developing methods involving computational machine learning—using algorithms to learn from data and make predictions—to study RA and other autoimmune diseases, drawing on large-scale clinical and preclinical single-cell datasets. That work, she says, could drive targeted interventions that could prevent the disease's progression.

"There's been significant research into how to treat a patient after someone is diagnosed," she says. "But there have been fewer studies into developing preventive strategies and identifying which healthy people are at risk of developing RA in the next couple of years. That's much more challenging. So we focus on enhancing disease prediction, ultimately enabling early disease prevention."

Bridging data science with translational medicine

RA is a chronic autoimmune disease, meaning it's a disorder in which the body's immune system mistakenly attacks its own healthy tissue, causing inflammation. Although RA is often associated with swelling, pain, and stiffness in the joints, it can affect various parts of the body, including the heart and lungs.

It's estimated that about 18 million people worldwide live with RA, 1.5 million of them in the United States. Nearly three times as many women have the disorder as men.

Available treatments can reduce inflammation and provide some relief, but there are no effective preventive treatments and no cures. The cause is uncertain, although RA has been associated with certain genes that may be triggered by a range of external factors.

Research has shown that many people who eventually develop RA symptoms experience immunological abnormalities that can be detected through blood tests years before the symptoms appear. Yet the length of this symptom-free "preclinical" phase can vary widely, and some people with these abnormalities never develop the full disease.

What's needed, Zhang says, are more precise ways to predict which people with preclinical abnormalities—or with a family history of RA—will progress to the full disease and how soon.

Zhang describes her work as a "bridge" between data science and translational medicine.

"Our research is very interdisciplinary," Zhang says. "We have large-scale data from patients with autoimmune disease, so that gives us the opportunity to apply our AI tools to various cohorts of patients."

Zhang's team analyses data on genetics, genomics, epigenetics, protein, and other factors from individual cells at various timepoints over long periods—known as single-cell multi-modal sequencing.

"Putting all these things together, we can hope to more robustly identify new and more accurate markers for prediction, combined with clinical characteristics," she says.

Pinpointing key immunological changes

The study presented in Zhang's new paper, titled "Deep immunophenotyping reveals circulating activated lymphocytes in individuals at risk for rheumatoid arthritis," has helped lay the foundation for her next phase of research.

Zhang's lab will apply their advanced computational tools to complex datasets collected from a large preclinical trial called StopRA. This, Zhang says, will strengthen her collaboration with CU rheumatologist Kevin Deane, MD, Ph.D., as they compare people who progressed to the disease with those who didn't. The goal is to pinpoint changes in the immune system associated with the progression from preclinical RA arthritis to symptoms.

In this publication, Zhang and her colleagues analysed RNA and protein expression in cells to compare people at risk of developing RA to those with symptoms as well as healthy people. They found "significant" differences in certain types of immune cells, particularly the expansion of specific T cell subtypes, in the at-risk group.

Those cells "could be a promising marker" for RA onset, Zhang says, and could lead to improved prevention strategies. But she says coming up with reliable markers is "still a ways off," and will require even larger and more geographically diverse datasets to see if the results she's seeing hold up.

Zhang is the corresponding author of this publication; her lab's postdoctoral fellow, Jun Inamo, MD, Ph.D., is the first author; and Deane and another rheumatology colleague, V. Michael Holers, MD, are among the co-senior authors.

More information: Jun Inamo et al, Deep immunophenotyping reveals circulating activated lymphocytes in individuals at risk for rheumatoid arthritis, Journal of Clinical Investigation (2025). DOI: 10.1172/JCI185217