Scientists at Mount Sinai have developed V2P, an artificial intelligence tool that identifies disease-causing genetic mutations and predicts the specific diseases they may trigger. The method uses advanced machine learning to link genetic variants with likely phenotypic outcomes, potentially accelerating genetic diagnostics and treatment discovery for complex and rare diseases.
Researchers at the Icahn School of Medicine at Mount Sinai have created a novel artificial intelligence system that goes beyond identifying harmful genetic mutations to predict the specific types of diseases these variants may cause. The breakthrough could significantly improve the speed and accuracy of genetic diagnosis whilst aiding drug discovery efforts for
rare and complex conditions.
The method, called Variant to Phenotype (V2P), addresses a critical gap in current genetic analysis. Whilst existing tools can estimate whether a mutation is harmful, they cannot determine what type of disease it might trigger. V2P uses advanced machine learning to connect genetic variants with their likely phenotypic outcomes essentially predicting how a patient’s DNA could influence their health across 23 top-level disease categories.
Streamlining variant interpretation
“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, PhD, who recently completed his doctoral training at Mount Sinai. “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 research team trained V2P on a large database containing both harmful and benign genetic variants, incorporating disease information to improve prediction accuracy. When tested using real, de-identified patient data, V2P consistently ranked the true disease-causing variant amongst the top 10 candidates, demonstrating its potential to streamline genetic diagnostics.
The tool’s applications extend beyond clinical diagnosis. According to co-senior and co-corresponding author Avner Schlessinger, PhD, Professor of Pharmacological Sciences and Director of the AI Small Molecule Drug Discovery Centre at Mount Sinai, “V2P could help researchers and drug developers identify the genes and pathways most closely linked to specific diseases. This can guide the development of therapies that are genetically tailored to the mechanisms of disease, particularly in rare and complex conditions.”
Performance validation
The authors report in their Nature Communications paper that V2P demonstrated superior performance compared to existing variant effect predictors across multiple evaluation datasets. For 21 of 22 phenotypes examined, V2P’s phenotype-specific predictions surpassed both its own general pathogenicity predictions and all compared methods. The tool achieved average precision scores of 0.86, 0.93, and 0.94 for separating pathogenic and benign variants across three distinct evaluation datasets.
Importantly, V2P showed strong concordance with experimental data from deep mutational scanning assays of 52 proteins, achieving correlation coefficients comparable to top-performing methods. In practical applications using patient sequencing data, V2P ranked causal variants significantly lower (median rank of two) compared to other genome-wide prediction methods.
Towards precision medicine
Whilst V2P currently classifies mutations into broad categories such as nervous system disorders or cancers, the research team aims to refine the tool to predict more specific disease outcomes and integrate it with additional data sources to support drug discovery.
“V2P gives us a clearer window into how genetic changes translate into disease, which has important implications for both research and patient care,” explains co-senior and co-corresponding author Yuval ltan, PhD, Associate Professor of Artificial Intelligence and Human Health, and Genetics and Genomic Sciences at Mount Sinai. “By connecting specific variants to the types of diseases they are most likely to cause, we can better prioritise which genes and pathways warrant deeper investigation.”
This innovation represents a step towards precision medicine, in which treatments can be matched to a patient’s genetic profile. The tool has been made available to the research community at www.v2p.ai, where scientists can access precomputed V2P scores for every possible single nucleotide variant in the human genome.
Reference:
Stein, D., Kars, M. E., Milisavljevic, B., et. al. (2025). Expanding the utility of variant effect predictions with phenotype-specific models.
Nature Communications, 16(1), 11113. https://doi.org/10.1038/s41467-025-66607-w
