Researchers at Mount Sinai have developed V2P, an AI tool that predicts which diseases specific DNA mutations are likely to cause, aiming to speed diagnosis and guide precision treatments.
A new artificial intelligence tool developed at the Icahn School of Medicine at Mount Sinai could help doctors move more quickly from a patient’s DNA sequence to answers about what is making them sick.
The method, called V2P (Variant to Phenotype), does more than flag potentially harmful genetic mutations. It also predicts the broad type of disease those mutations are likely to cause, offering a way to connect changes in DNA to real-world symptoms.
The work, published in the journal Nature Communications, is designed to speed up genetic diagnostics and support the search for new treatments for complex and rare conditions.
Today, when a patient has their genome or exome sequenced, clinicians often end up with a list of thousands of genetic variants. Existing tools can estimate whether a given variant is likely to be harmful, but they typically stop there. They do not indicate whether a mutation is more likely to lead to a nervous system disorder, a cancer, or another category of disease.
V2P is meant to fill that gap. The system uses machine learning to link specific genetic variants to their likely phenotypic outcomes, meaning the diseases or traits that might result. In practice, that could help clinicians focus on the handful of variants most likely to explain a patient’s symptoms.
The approach is designed to cut through the noise of large genetic datasets, according to first author David Stein, who recently completed his doctoral training in the labs of Yuval Itan and Avner Schlessinger.
“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,” Stein said in a news release. “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.”
To build V2P, the team trained the model on a large database of genetic variants that included both harmful and benign changes, along with information about associated diseases. By learning from examples where the outcome is already known, the AI can then make predictions about new variants it has not seen before.
The researchers tested V2P on real, de-identified patient data. In those tests, the tool often ranked the true disease-causing variant among the top 10 candidates. That kind of performance, the team says, could streamline the diagnostic process by helping clinicians zero in on the most promising leads sooner.
While the current version of V2P classifies mutations into broad categories such as nervous system disorders or cancers, the researchers see this as a starting point. They aim to refine the tool so it can predict more specific disease outcomes and integrate it with additional data sources, such as information about how genes interact or how proteins function in cells.
Beyond helping individual patients, the team believes V2P could be a powerful resource for researchers and drug developers.
“Beyond diagnostics, V2P could help researchers and drug developers identify the genes and pathways most closely linked to specific diseases,” added Schlessinger, a co-senior and co-corresponding author, professor of pharmacological sciences and director of the AI Small Molecule Drug Discovery Center at the Icahn School of Medicine at Mount Sinai. “This can guide the development of therapies that are genetically tailored to the mechanisms of disease, particularly in rare and complex conditions.”
That kind of insight is central to the broader push toward precision medicine, an approach that aims to match treatments to a person’s unique genetic and biological profile. By clarifying which genetic changes are most likely to drive a disease, tools like V2P could help scientists identify new drug targets and design therapies that act on the right pathways.
The technology offers a more direct view of how DNA changes affect health, according to Itan, a co-senior and co-corresponding author and an associate professor of artificial intelligence and human health and of genetics and genome sciences at the Icahn School of Medicine at Mount Sinai.
“V2P gives us a clearer window into how genetic changes translate into disease, which has important implications for both research and patient care,” he said in the news release. “By connecting specific variants to the types of diseases they are most likely to cause, we can better prioritize which genes and pathways warrant deeper investigation. This helps us move more efficiently from understanding the biology to identifying potential therapeutic approaches and, ultimately, tailoring interventions to an individual’s specific genomic profile.”
The study highlights how adding disease-specific information to existing prediction tools can expand their usefulness. Instead of treating all harmful variants as essentially the same, V2P attempts to distinguish how different mutations might lead to different categories of illness.
For patients with rare or unexplained conditions, this kind of technology could be especially valuable. Many such patients endure years of testing and specialist visits before receiving a diagnosis, a journey often called a diagnostic odyssey. A tool that can prioritize the most likely disease-causing variants and suggest the type of disorder involved could help shorten that path.
The researchers plan to continue improving V2P’s accuracy and resolution, with the goal of moving from broad disease categories to more precise predictions. They also envision integrating the tool into research pipelines for drug discovery, where it could help identify promising molecular targets earlier in the process.
As genomic sequencing becomes more common in clinics and research labs, the challenge is shifting from collecting DNA data to making sense of it. V2P represents one attempt to bridge that gap, using artificial intelligence to translate raw genetic information into insights that could ultimately improve diagnosis and treatment.