When 12-year-old Lila from Melbourne finally received a diagnosis for her unexplained developmental delays, it wasn’t from a new test—but from a second look at her old DNA data, powered by artificial intelligence. Her story is one of 241 newly solved medical mysteries, thanks to Talos, an open-source AI tool developed by researchers across Australia and the United States. For families long caught in diagnostic limbo, Talos is delivering answers that were once out of reach.
Rare diseases, though individually uncommon, collectively affect more than one in 17 people in Australia. While genomic sequencing has transformed diagnosis, over half of patients leave initial testing without answers. The problem isn’t just the complexity of the data—it’s the pace of science. New gene-disease links emerge monthly, but manually reanalyzing old genomes is too slow and expensive for most clinics. That’s where Talos steps in.
Created by a collaboration led by the Murdoch Children’s Research Institute (MCRI), Victorian Clinical Genetics Services (VCGS), the Centre for Population Genomics, the Broad Institute of MIT and Harvard, and Microsoft Research, Talos automates the reanalysis of genomic data, integrating the latest scientific discoveries every month. In a validation study of 1,089 patients, it correctly identified nearly 90% of known diagnoses while flagging just 1.3 candidate variants per family—making it both accurate and efficient. When applied to 4,735 previously undiagnosed patients, Talos delivered 241 new diagnoses, boosting diagnostic yield by 5.1% across conditions including neurodevelopmental, cardiac, and renal disorders.
More than half of these diagnoses were possible only because of newly published scientific knowledge. Remarkably, Talos made these connections fast: the median time from knowledge release to diagnosis was just 32 days, with some answers surfacing in as little as one day. The cost? Less than USD$12 to run the initial analysis on 1,000 genomes, and under USD$2 per year for monthly reanalysis. This efficiency opens the door to routinely updating diagnoses across entire patient populations—not just the few who can access specialized centers.
The impact extends beyond the individual. In over 50 family members, new genetic insights have already led to preventive screenings, tailored treatments, and informed reproductive choices—especially for inherited heart conditions that could otherwise strike without warning. As Professor Zornitza Stark of MCRI puts it, “We developed Talos to overcome these barriers by automating the reanalysis process.” With genomic data only growing in healthcare, tools like Talos ensure that every genome keeps working for the patient, long after the initial test.