In a groundbreaking development, researchers have unveiled an artificial intelligence (AI) tool capable of diagnosing a range of diseases — including diabetes, HIV, COVID-19, and autoimmune conditions — from a single blood sample. This innovative “one-shot” diagnostic method, detailed in The Nature and published in Science on February 20, could transform the way overlapping conditions are detected and treated.
The AI system, developed by a collaborative team from the University of Cambridge and Stanford University, analyzes immune cell gene sequences to provide a comprehensive snapshot of a person’s health. According to lead researcher Sarah Teichmann, a molecular biologist at Cambridge, the tool captures “everything that your immune system has been exposed to,” offering unprecedented diagnostic potential.
While the technology is not yet ready for clinical use, co-author Maxim Zaslavsky, a computer scientist at Stanford, highlighted its future promise: “This could help diagnose conditions that currently lack definitive tests, offering clinicians a powerful tool to manage complex diseases.”
Decoding the Body’s Natural Diagnostic System
The immune system maintains a detailed record of past infections and illnesses through two primary cell types: B cells and T cells. B cells produce antibodies that combat infections, while T cells are responsible for killing infected cells or activating other immune responses. By analyzing gene sequences that encode these cells’ receptors, researchers can identify patterns that reflect both past and present diseases.
Zaslavsky and his team created an AI tool that integrates six machine-learning models to analyze 16.2 million B-cell receptors and 23.5 million T-cell receptors from nearly 600 participants. The study sample included individuals with COVID-19, HIV, lupus, type 1 diabetes, and those recently vaccinated against the flu, alongside healthy controls.
The tool demonstrated impressive accuracy, scoring 0.986 on a metric measuring its ability to correctly identify participants’ health conditions—a near-perfect performance. The study also revealed that combining B-cell and T-cell data provided the most accurate results. For instance, autoimmune diseases like type 1 diabetes and lupus had clearer signatures in T-cell receptors, while infections like COVID-19 and HIV were more distinguishable through B-cell receptors.
The Road to Clinical Application
Despite the promising results, researchers caution that the technology requires further refinement before it can be used in clinical settings. According to Victor Greiff, a computational immunologist at the University of Oslo, the tool must outperform current diagnostic methods to become a standard medical practice.
The AI tool’s predictive abilities could also open doors to personalized treatments. Scott Boyd, an immunologist from Stanford University School of Medicine, suggests that analyzing cases where the AI misclassified diseases could help reveal previously undetected subcategories of immunological conditions. This level of detail could allow clinicians to tailor therapies more effectively to individual patients.
