What You Should Know:
– Scientists at the Icahn School of Medicine at Mount Sinai have developed a new artificial intelligence tool called V2P (Variant to Phenotype) that identifies not only disease-causing genetic mutations but also predicts the specific diseases they may trigger.
– Published in Nature Communications, this machine learning model improves diagnostic speed and accuracy by linking genetic variants to phenotypic outcomes, moving beyond simple “harmful vs. benign” classifications. The tool aims to streamline genetic interpretation for clinicians while guiding drug developers toward genetically tailored therapies for rare and complex conditions.
Beyond “Harmful”: Mount Sinai’s V2P AI Predicts Which Disease Your Genes Might Cause
For years, genetic testing has faced a “last mile” problem. We can sequence a genome and identify thousands of variants, and we can often tell if a variant is “harmful.” But knowing that a mutation is bad is very different from knowing what it will do. Today, researchers at the Icahn School of Medicine at Mount Sinai announced a solution that bridges that gap.
Their new AI tool, V2P (Variant to Phenotype), is designed to predict the specific disease or trait a genetic mutation is likely to trigger. The findings, published today in Nature Communications, suggest that V2P could fundamentally change the speed and accuracy of diagnosing rare disorders.
Solving the Specificity Gap
Current computational tools for genetics are largely binary: they estimate whether a variant is pathogenic (disease-causing) or benign. They rarely offer context. This leaves clinicians with a list of “potentially harmful” mutations but no roadmap to the patient’s actual condition. V2P was built to provide that context. By using advanced machine learning, the tool links genetic variants directly to their phenotypic outcomes—the observable traits or diseases they produce.
“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,” said David Stein, PhD, the study’s first author. “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”.
How It Works
The model was trained on a massive database of known harmful and benign variants, integrating disease-specific information to “teach” the AI the relationship between code and condition.
In validation tests using de-identified patient data, V2P demonstrated high accuracy, often ranking the true disease-causing variant among the top 10 candidates. For a clinician faced with a mysterious set of symptoms and a complex genetic report, this ranking capability acts as a high-powered filter, drastically reducing the time required to reach a diagnosis.
Implications for Drug Discovery
The potential of V2P extends beyond the clinic and into the lab. Dr. Avner Schlessinger, Director of the AI Small Molecule Drug Discovery Center at Mount Sinai, views the tool as a compass for drug development.
“V2P could help researchers and drug developers identify the genes and pathways most closely linked to specific diseases,” Schlessinger noted. “This can guide the development of therapies that are genetically tailored to the mechanisms of disease, particularly in rare and complex conditions”.
By clarifying the biological mechanisms driven by specific variants, V2P helps scientists prioritize which genetic pathways warrant deeper investigation, moving the industry closer to true precision medicine.
The Road Ahead
Currently, V2P classifies mutations into broad disease categories, such as nervous system disorders or cancers. The research team, led by co-senior author Dr. Yuval Itan, plans to refine the tool to predict even more granular disease outcomes.
Future iterations will integrate additional data sources to further support drug discovery. “V2P gives us a clearer window into how genetic changes translate into disease,” said Dr. Itan. “This helps us move more efficiently from understanding the biology to identifying potential therapeutic approaches”.
