A newly developed AI tool can speed up the search for the genetic causes of rare diseases, a process that often takes years and frequently ends without answers.
The tool analyses how genes have evolved across many species to uncover hidden clues about which gene is responsible for a patient’s symptoms. In tests, it successfully identified the disease-causing gene in most cases, even when that gene had never previously been linked to a disease. This approach could significantly shorten the diagnostic journey and help guide doctors toward effective treatments much sooner.
For families of children living with rare diseases, the search for a diagnosis is often long, uncertain, and exhausting, marked by years of unanswered questions, repeated tests, and the absence of clear medical explanations. This journey can last nearly a decade. Even in the age of advanced genetic sequencing, most patients remain undiagnosed.
A new study led by Christina Canavati and Yuval Tabach from the Faculty of Medicine at the Hebrew University of Jerusalem, published in Genetics in Medicine, may change that. It builds on earlier research conducted with Nobel Prize laureate Gary Ruvkun, published in 2013.
At the heart of the breakthrough is EvORanker, an AI algorithm designed to enable identifying which gene, among thousands of possibilities, is actually causing a patient’s disease.
Instead of relying only on existing medical knowledge, EvORanker looks across evolution. By comparing genetic patterns across more than 1,000 species, the algorithm detects hidden relationships between genes, even those that science has never linked to disease before.
In clinical testing, the algorithm identified the correct disease-causing gene as the top candidate in nearly 70% of cases, and placed it within the top five in 95% of cases, outperforming existing tools, especially in the most challenging scenarios involving poorly understood genes.
In one case described in the study, a child with a complex neurodevelopmental disorder had undergone extensive testing without a diagnosis. Using EvORanker, researchers identified a previously unrecognised gene as the likely cause, opening the door to understanding the disease and, potentially, treating it.
In another case, the algorithm revealed the genetic basis of a severe multisystem disorder affecting multiple organs. The discovery not only provided answers to the family but also pointed researchers toward possible therapeutic strategies.
“These are thousands of cases like that around the world that fall through the cracks of current medicine,” Tabach said.
“Our goal was to give patients and clinicians a tool, that can find fast and accurate answers where none existed before.”
By uncovering new disease genes, EvORanker also helps identify existing drugs that could be repurposed, a shortcut that could save years of development time and bring treatments to patients faster.
The research builds on more than a decade of work combining evolutionary biology and computational science. Earlier discoveries by Tabach and collaborators demonstrated how genes that evolve together often function together, this leads to dozens breakthrough. EvORanker turns that principle into a diagnostic engine.
And while rare diseases are the immediate focus, the team is already looking ahead. The technology is now being applied to cancer, where researchers are using it to uncover why some tumours unexpectedly regress, and how those mechanisms could be harnessed for treatment of stage 4 cancer patients.
EvORanker is now available as an accessible tool for researchers and clinicians, with additional studies already underway and new clinical applications emerging.