Researchers at the Icahn School of Medicine at Mount Sinai:
have developed a novel artificial intelligence framework capable of predicting phenotype-specific disease outcomes from genetic variants. Unlike traditional variant analysis tools, this system goes beyond pathogenicity assessment to forecast the type of disease a mutation is likely to cause.
The system, known as V2P (Variant to Phenotype),
represents a significant advancement in genetic interpretation, with implications for clinical diagnostics, precision medicine, and drug discovery. The findings were published on December 15, 2025, in Nature Communications.
Addressing a Key Limitation in Genetic Interpretation:
Current variant effect prediction tools are primarily designed to assess whether a genetic mutation is damaging. However, they typically do not provide insight into downstream clinical consequences. As a result, clinicians must manually correlate genetic findings with patient symptoms, a process that is both time-consuming and error-prone. V2P addresses this gap by leveraging machine learning to map genetic variants directly to disease phenotypes. This enables the prioritization of mutations based not only on pathogenicity, but also on their likely clinical relevance.
Model Training and Performance:
The V2P model was trained using a large, curated dataset containing benign and pathogenic variants paired with detailed disease annotations. Through this process, the system learned associations between specific genetic alterations and phenotypic outcomes. Validation using real-world, de-identified patient data demonstrated that V2P frequently ranked the true disease-causing variant within the top ten candidates, significantly improving diagnostic efficiency.
According to first author Dr. David Stein, this approach allows clinicians and researchers to narrow their focus to variants that are most likely to explain a patient’s condition, improving both speed and accuracy of interpretation.
Implications for Drug Discovery and Precision Medicine:
Beyond clinical diagnostics, V2P offers value in identifying disease-relevant genes and pathways, providing actionable insights for therapeutic development. By highlighting genetically driven mechanisms, the system supports the design of targeted treatments, particularly for rare and multifactorial diseases.
Dr. Avner Schlessinger, co-senior author and Director of the AI Small Molecule Drug Discovery Center at Mount Sinai, emphasized that phenotype-specific predictions can guide drug development strategies toward genetically validated targets. Future Directions Currently, V2P classifies variants into broad disease categories, including cancer and neurological disorders. Ongoing work aims to increase prediction granularity and integrate additional biological datasets to enhance performance.
As emphasized by Dr. Yuval Itan, co-senior author of the study, linking genetic variation to phenotype-specific outcomes accelerates the transition from genomic data to therapeutic intervention, reinforcing the foundations of precision medicine.
Conclusion:
V2P represents a meaningful advance in AI-driven genomics by transforming variant analysis into clinically actionable insight. By predicting disease outcomes directly from DNA mutations, this system moves healthcare closer to truly personalized diagnosis and treatment.
