A new artificial intelligence system is helping doctors diagnose rare genetic diseases in days instead of years, marking a significant leap forward for patients who often face long and uncertain diagnostic journeys. The tool, which analyzes patient DNA and medical records simultaneously, has already identified conditions in hundreds of undiagnosed cases during early testing.
The system, called Fabric GEM, works by scanning a patient’s entire genome for disease-causing variants and cross-referencing them with electronic health records. In a study of over 1,000 patients with suspected genetic disorders, the AI correctly identified the underlying cause in 68 percent of cases that had previously stumped human experts. For patients with rare diseases, the average time to diagnosis dropped from more than five years to under one month.
The technology is particularly effective for conditions that are difficult to spot because they mimic more common illnesses. The AI can detect patterns across thousands of genes and symptoms simultaneously, something that would take a human geneticist weeks or months to complete. The tool is now being deployed in several major hospital systems, where it is helping to diagnose conditions including muscular dystrophy, metabolic disorders, and severe childhood epilepsy syndromes.
How It Changes Patient Care
For patients and families, the speed of diagnosis can be life-changing. Many rare disease patients spend years visiting multiple specialists, undergoing repeated tests, and receiving incorrect treatments before finally learning the cause of their symptoms. With the AI system, doctors can order a single test that returns a diagnosis within days, allowing them to begin targeted therapy or connect families with support networks immediately.
The tool also helps identify patients who may be eligible for clinical trials or gene therapies that are specific to their mutation. In the study, approximately 12 percent of patients who received a diagnosis through the AI were found to have a variant that could potentially be treated with an existing or experimental therapy.
Researchers are now working to expand the system to include more diverse populations, since most genetic databases have historically been built from people of European ancestry. Early results from a pilot program in Africa show the AI can maintain its accuracy across different ethnic groups when trained on representative data. The team behind the tool plans to make it available to more hospitals worldwide over the next year, with a particular focus on regions where access to genetic specialists is limited.