Mount Sinai researchers develop Variant to Phenotype tool that identifies harmful mutations and predicts disease types, accelerating diagnosis and drug discovery
Scientists have unveiled an artificial intelligence tool that transforms how doctors identify disease-causing genetic mutations by not only detecting harmful DNA changes but also predicting which specific diseases those mutations will likely trigger. The breakthrough, developed at the Icahn School of Medicine at Mount Sinai, promises to dramatically accelerate genetic diagnosis while opening new pathways for developing treatments targeting rare and complex conditions.
The innovative method, named Variant to Phenotype or V2P, addresses a critical limitation in current genetic analysis by creating direct connections between DNA alterations and their probable disease outcomes. This advancement represents a significant leap beyond existing genetic testing capabilities that leave medical professionals struggling with incomplete information.
Why current genetic testing falls short
Existing genetic testing tools perform well at estimating whether a particular genetic variant might cause harm, but their usefulness essentially ends at that determination. Clinicians frequently find themselves confronting extensive lists of possible mutations without clear guidance about which ones actually relate to their patient’s symptoms or condition.
This gap creates frustrating delays and uncertainty in the diagnostic process. Doctors must manually sift through potentially thousands of genetic variants, attempting to identify which changes matter most for their specific patient. The process consumes valuable time while patients wait anxiously for answers about their health conditions.
V2P was specifically designed to bridge this knowledge gap using sophisticated machine learning techniques. The system predicts not only whether a variant poses danger but also categorizes the type of disease it will most likely cause, whether neurological disorders, cancer or other specific disease categories.
Focusing on what matters most for patients
David Stein, the study’s first author, explained how the new approach fundamentally changes genetic interpretation. Rather than forcing medical teams to examine thousands of possible variants without direction, the tool pinpoints genetic changes most relevant to each patient’s specific condition.
By determining both the pathogenic nature of a variant and the disease type it will likely trigger, V2P improves both speed and accuracy in genetic interpretation and diagnostics. This dual capability represents a meaningful advancement over single-function tools that only identify harmful mutations without providing disease context.
Teaching the AI to recognize patterns
The research team trained V2P using an extensive dataset combining both harmful and benign genetic variants alongside detailed disease information. This comprehensive training enabled the artificial intelligence model to identify patterns connecting specific mutations to particular phenotypic outcomes.
Testing on real, de-identified patient data demonstrated impressive performance. The tool successfully ranked the true disease-causing mutation among the top 10 candidates, suggesting V2P could substantially streamline diagnostic processes in clinical genetics settings. This accuracy level means doctors can quickly narrow their focus to the most relevant genetic changes rather than investigating dozens or hundreds of possibilities.
Transforming drug development beyond diagnosis
While improved diagnostics represent an obvious benefit, researchers believe V2P’s applications extend significantly into biomedical research and pharmaceutical development. The tool helps scientists and drug developers identify genes and pathways most closely connected to specific diseases, creating more targeted research directions.
Avner Schlessinger, co-senior and co-corresponding author, professor of pharmacological sciences and director of the AI Small Molecule Drug Discovery Center at Mount Sinai, emphasized how this capability guides development of therapies genetically tailored to disease mechanisms. The approach proves particularly valuable for rare and complex conditions where treatment options remain limited.
Current pharmaceutical development often proceeds through trial and error, testing compounds against diseases without fully understanding the genetic mechanisms involved. V2P accelerates this process by identifying which genes and pathways deserve focused investigation, potentially saving years of research time and millions of dollars in development costs.
Building toward more precise predictions
At present, V2P classifies mutations into broad disease categories rather than highly specific diagnoses. The development team plans to refine the system’s capabilities so it can predict more granular disease outcomes with greater precision.
Future versions will integrate additional biological data to further support drug discovery efforts. This expansion will create even more detailed connections between genetic variants and disease manifestations, providing researchers with increasingly specific targets for therapeutic intervention.
Advancing personalized medical care
The research team describes V2P as an important step toward true precision medicine, where both diagnosis and treatment align closely with each individual’s unique genetic profile. This personalized approach represents a departure from one-size-fits-all medicine that has dominated healthcare for generations.
Yuval Itan, co-senior and co-corresponding author and associate professor of artificial intelligence, human health, genetics and genomic sciences at Mount Sinai, explained how V2P creates a clearer window into how genetic changes translate into disease. This understanding carries important implications for both research advancement and patient care improvement.
By connecting specific variants to the disease types they most likely cause, medical teams can better prioritize which genes and pathways warrant deeper investigation. This prioritization helps the field move more efficiently from understanding underlying biology to identifying potential therapeutic approaches.
Ultimately, this progression leads to tailoring interventions to each individual’s specific genomic profile rather than relying on treatments developed for average patients. The shift toward genomically-informed medicine promises more effective treatments with fewer side effects, as therapies target the precise genetic mechanisms causing disease in each person.
The tool represents meaningful progress in transforming genetic information from abstract data into actionable medical intelligence that directly improves patient outcomes.
SOURCE: Drug TARGET REVIEW