What if doctors could look at a person’s DNA and not only spot a harmful mutation, but also predict the kind of disease it may cause—before symptoms fully appear? Scientists at the Icahn School of Medicine at Mount Sinai say that the future is closer than ever.
They have developed a new artificial intelligence tool, V2P (Variant to Phenotype), that connects genetic mutations directly to the diseases they are likely to trigger. The advance could transform how rare and complex illnesses are diagnosed and treated. The findings are published in Nature Communications.
Genetic testing has long helped doctors identify mutations that may be dangerous. But in most cases, clinicians are left guessing what those mutations actually mean for a patient’s health. V2P changes that, say the authors. Using machine learning, the tool predicts not just whether a genetic change is harmful, but the type of disease it is most likely to cause—such as a neurological disorder or cancer.
"Our approach allows us to pinpoint the genetic changes that are most relevant to a patient’s condition, rather than sifting through thousands of possible variants," says first author David Stein, who recently completed his doctoral training.
"By determining not only whether a variant is pathogenic but also the type of disease it is likely to cause, we can improve both the speed and accuracy of genetic interpretation and diagnostics."
The tool was trained on large datasets containing both harmful and harmless genetic variants, along with detailed disease information. When tested on real patient data, V2P consistently ranked the true disease-causing mutation among the top candidates, reducing the long and costly search often faced by families with rare diseases.
The implications go beyond diagnosis. By linking genes to specific disease pathways, V2P can also guide drug discovery. Researchers can identify which biological processes are most important in a disease, helping scientists design treatments that target the root cause rather than just symptoms.
“This approach can point drug developers toward the genes and pathways that matter most,” adds Dr. Avner Schlessinger, a senior author of the study. “That is especially important for rare diseases, where treatment options are limited or nonexistent.”
For now, V2P predicts disease categories rather than exact conditions. But the team aims to refine the system to make more precise predictions and integrate it with other biological data, bringing medicine closer to truly personalised care.
Experts say the study marks a key step toward precision medicine, where treatment decisions are guided by a patient’s unique genetic profile. By translating genetic data into meaningful clinical insight, tools like V2P could shorten diagnostic delays, reduce uncertainty, and open new paths for targeted therapies.
“Understanding how genetic changes lead to specific diseases is crucial,” says Dr. Yuval Itan, a co-senior author. “This tool helps us connect the dots between DNA and disease—and that can change both research and patient care.”
The other authors of the paper titled “Expanding the utility of variant effect predictions with phenotype-specific models,” included Meltem Ece Kars, Baptiste Milisavljevic, Matthew Mort, Peter D. Stenson, Jean-Laurent Casanova, David N Cooper, Bertrand Boisson, and Peng Zhang.
