Researchers have developed a new AI algorithm called EvORanker that dramatically speeds up the search for genetic causes of rare diseases, a process that has traditionally taken years and often ends without a diagnosis. By comparing genetic evolutionary patterns across more than 1,000 species, the tool identified a previously unrecognized gene as the likely cause of a rare disease case, according to a 2026 report from the American Friends of Hebrew University.
Why Rare Disease Genetics Is So Hard
An estimated 7,000 rare diseases affect roughly 30 million Americans, and for many patients, geneticists know something is wrong at the molecular level but cannot pinpoint which specific gene is responsible. Traditional gene-disease discovery relies heavily on finding multiple unrelated patients with the same rare condition who happen to share mutations in the same gene, a statistical approach that simply does not work when a disease is rare enough that only a handful of cases exist worldwide.
How EvORanker Works
EvORanker takes a fundamentally different approach by looking at evolutionary conservation patterns, essentially asking how a gene’s function and structure have changed, or stayed stable, across more than 1,000 different species over evolutionary time. Genes that are highly conserved across distantly related species tend to be doing something biologically essential, and unusual patterns in that conservation can reveal hidden functional relationships between genes that medical research has never previously connected to a specific disease. This method allows researchers to flag candidate genes even when there are too few patient cases to rely on traditional statistical gene-disease association methods.
A Concrete Success Story
In the case highlighted by researchers, EvORanker identified a gene that had never before been linked to the rare disease in question as the likely underlying cause. That identification, according to the report, opens the door not just to a diagnosis for affected patients but potentially to future treatment research, since knowing the specific gene involved gives scientists a target to study and, eventually, try to correct or compensate for therapeutically.
Part of a Broader AI Push in Rare Disease Diagnosis
EvORanker’s debut comes alongside other notable 2026 developments in AI-assisted rare disease diagnosis, including a Boston Children’s Hospital and OpenAI collaboration that used a reasoning-focused AI model to help confirm 18 new diagnoses among 376 previously unsolved pediatric cases. Together, these efforts reflect a broader recognition within the genetics research community that rare diseases, precisely because they are individually uncommon but collectively affect tens of millions of people, are a category where AI’s pattern-recognition capabilities can add genuine value that traditional statistical methods cannot easily replicate.
Caveats From the Genetics Community
Geneticists caution that evolutionary conservation analysis, while powerful, is one tool among many needed to confirm a gene-disease link, and that any AI-flagged candidate gene still requires substantial laboratory validation, including functional studies and additional patient data, before it can be considered a confirmed diagnosis. There is also the broader challenge, common across AI diagnostic tools, of ensuring these algorithms perform reliably across genetically diverse populations rather than reflecting biases present in existing genomic reference databases, which have historically underrepresented non-European ancestries.
What’s Next
Researchers behind EvORanker say they plan to apply the algorithm to additional unsolved rare disease cases, building on the initial proof-of-concept success. As genomic sequencing becomes cheaper and more widely used in clinical practice, tools like EvORanker that can extract more diagnostic value from existing genetic data, without requiring additional testing, are likely to become an increasingly important part of the rare disease diagnostic toolkit.
Building on Decades of Comparative Genomics
EvORanker’s approach builds on decades of comparative genomics research showing that genes essential to basic biological functions tend to be conserved across widely divergent species, from yeast to humans, a principle scientists have used for years in laboratory research but rarely applied at the scale of more than 1,000 species simultaneously. Researchers say the computational power now required to run this kind of large-scale cross-species comparison efficiently has only become practical recently, which helps explain why tools like EvORanker are emerging now rather than a decade ago when the underlying evolutionary biology concept was already well understood. Researchers say EvORanker’s underlying comparative genomics database will continue to grow as more species genomes are sequenced and added to public repositories, meaning the tool’s diagnostic power is likely to improve incrementally with each new genome added to its reference set.
