EvORanker: the AI Revolutionizing Rare Disease Diagnosis

A Hope for Families Seeking Diagnosis

EvORanker, an algorithm developed at the Hebrew University of Jerusalem, could be a game changer for families facing rare diseases. Published in Genetics in Medicine, this tool has demonstrated its ability to rank the gene responsible for a rare disease first in 69% of tested cases, and in the top 5 in 95% of cases. These results surpass those of existing tools, particularly for lesser-known genes.

The algorithm compares the genetic profiles of over 1,028 eukaryotic genomes to identify genes that have co-evolved. This method is based on the idea that genes evolving together tend to function together in the organism. By using an evolutionary signal combined with a protein interaction network from the STRING database, EvORanker is able to highlight genes often overlooked by other methods.

A Diagnosis That Is Often Long and Uncertain

For more than a quarter of patients with a rare disease, the wait before even starting the search for a diagnosis can stretch to nearly four years. Once this search begins, it lasts at least another year and a half for the majority of patients. The major issue is not the absence of genetic data, but rather the inability to determine which variant among the thousands identified should be examined first.

Christina Canavati and Yuval Tabach, researchers at the Hebrew University of Jerusalem, published the results of EvORanker on March 14, 2026. This AI-powered tool compares the evolution of genes across more than 1,000 species to rank the suspects by likelihood.

A Prioritization Tool, Not a Final Diagnosis

In the study, several families with children suffering from a severe neurodevelopmental syndrome had been waiting for a diagnosis for years. This syndrome affects the nervous system, skeleton, and transcription machinery. EvORanker identified the gene SUPT4H1 as a priority candidate. Follow-up analyses confirmed that biallelic variants in this gene cause a previously unreported disorder, characterized by dystonia, intellectual disability, and dental anomalies.

Yuval Tabach stated, “There are thousands of cases like this around the world that slip through the cracks of current medicine. Our goal was to provide patients and clinicians with a tool capable of finding answers where none existed.”

However, ranking SUPT4H1 at the top of the list is not sufficient to diagnose these children. After prioritization by EvORanker, it was necessary to verify that the identified variant was indeed pathogenic, confirm its presence in other family members, conduct functional modeling on the C. elegans worm, and cross-reference transcriptomic and proteomic data. This process involved teams in Jerusalem, Oslo, and Barcelona, a poorly developed infrastructure in hospitals.

Recommendations and Future Perspectives

The recommendations published in October 2025 in Genome Medicine on optimizing Exomiser and Genomiser emphasize that performance directly depends on the quality of symptoms recorded according to the standardized HPO vocabulary. Vague or incomplete terms can diminish the relevance of the candidate list. The reported 69% and 95% figures apply to well-documented data under controlled research conditions. Ongoing studies will need to establish whether these rates hold true in multi-ethnic cohorts, with partial records and without specialized geneticists.

The tool is available for researchers and clinicians. The team is working on its application to certain cancers, particularly to understand why tumors regress unexpectedly. This avenue was mentioned in the works of Yuval Tabach, which have already been co-published with Nobel laureate Gary Ruvkun since 2013.