Diagnosing autoimmune diseases has always required multiple steps: imaging, blood panels, biopsies, specialist referrals. Each test narrows the field, but misdiagnosis is common and diagnostic delays can stretch to years. A study published in Science in early 2025 suggests a fundamentally different approach may be possible — reading a disease's identity directly from the immune system's own record-keeping.
The study, led by Scott Boyd's group at Stanford, analyzed the T cell receptor (TCR) and B cell receptor (BCR) sequences present in whole blood from patients across more than a dozen autoimmune and immune-related diseases, including lupus, multiple sclerosis, rheumatoid arthritis, and type 1 diabetes. These receptors are generated uniquely in each immune cell and reflect what that cell has been trained to respond to — essentially a molecular log of the immune system's history.
The researchers found that an AI trained on the receptor sequence data could differentiate between diseases with high accuracy, even from bulk sequencing data rather than single-cell analysis. That's the surprising part. Most researchers assumed the signal would be too diluted in bulk samples — that you would need to isolate individual cells to find meaningful patterns. The results showed that wasn't the case.
What the model was picking up on is that different autoimmune conditions leave different fingerprints in the T cell repertoire. The immune system responds differently to each disease, and those differences are detectable even in a standard blood draw.
The implications for diagnostics are substantial. A single blood test, analyzed by AI, might one day replace the lengthy diagnostic odyssey that many autoimmune patients currently endure. The approach could also be used to monitor disease activity over time or to predict how a patient will respond to a specific treatment.
The team is working to validate the approach across broader patient populations and to refine the model's ability to distinguish diseases that share overlapping immune signatures. But the proof of concept is now firmly established: the immune system is keeping records, and we're getting better at reading them.