AI Breakthrough Predicts Disease Risk from Genetic Mutations with Unprecedented Accuracy
For years, the promise of personalized medicine has hinged on our ability to translate genetic information into actionable health insights. But interpreting the impact of rare genetic variations – understanding if and when a mutation will actually lead to disease – has remained a important challenge. Now, researchers at the Icahn School of medicine at Mount Sinai have unveiled a groundbreaking AI-powered method that dramatically improves our ability to assess this risk, moving us closer to truly proactive healthcare.
(Image credit: Icahn School of Medicine at Mount Sinai)
The Challenge of Genetic Penetrance
The core of this advancement lies in addressing a concept called “penetrance.” Simply put, penetrance refers to the proportion of individuals with a specific genetic mutation who actually develop the associated disease. Its rarely a simple yes or no. Many common conditions – think high blood pressure, diabetes, or various cancers – exist on a spectrum, making customary genetic analysis insufficient.
Historically, genetic studies have often relied on binary classifications. This new approach, detailed in the August 28th issue of Science, leverages the power of artificial intelligence to quantify disease risk along that spectrum, offering a far more realistic and nuanced understanding.
A Scalable, Accessible Solution
What sets this research apart is its practicality. instead of relying on complex and expensive specialized testing, the mount Sinai team trained AI models using readily available data: routine lab tests already included in most electronic health records (EHRs). Cholesterol levels, blood counts – information doctors routinely collect – become powerful predictors when analyzed through this AI lens.
“this is a much more nuanced, scalable, and accessible way to support precision medicine,” explains Dr. Ron Do, senior author of the study and the Charles Bronfman Professor in personalized Medicine at Mount Sinai. “It’s notably valuable when dealing with rare or ambiguous genetic findings.”
How the AI Works: ML Penetrance Scores
The researchers analyzed over 1 million EHRs to build AI models for 10 common diseases. These models were then applied to individuals carrying rare genetic variants, generating a “ML penetrance” score ranging from 0 to 1.
A score closer to 1: Indicates a higher likelihood that the variant contributes to disease advancement.
A score closer to 0: suggests minimal or no risk associated with the variant.
The results were often surprising.Variants previously flagged as “uncertain” revealed clear disease signals, while others believed to be highly impactful showed little effect in real-world patient data. This highlights the limitations of relying solely on theoretical predictions.
Empowering Clinical Decision-Making
Lead author Dr. Iain S. Forrest emphasizes that this AI model isn’t intended to replace clinical judgment. Instead, it’s designed to be a powerful guide, particularly when test results are inconclusive.
Consider Lynch syndrome, an inherited condition that increases cancer risk. A high ML penetrance score could prompt earlier and more frequent cancer screenings.Conversely, a low score could alleviate needless anxiety and avoid potentially harmful overtreatment.
Looking Ahead: Expansion and Long-Term Tracking
The Mount Sinai team isn’t stopping here. They are actively working to:
Expand the model: Include more diseases and a broader range of genetic changes.
Increase diversity: Ensure the model performs accurately across diverse populations.
* Longitudinal studies: Track the predictive power of the model over time and assess whether early interventions based on these predictions improve patient outcomes.
This research,supported by grants from the National Institutes of Health (NIH),represents a significant leap forward in our ability to translate genetic information into personalized healthcare.It’s a testament to the power of AI to unlock the complexities of the human genome and empower both clinicians and patients with the knowledge they need to make informed decisions about their health.
Funding Sources: National Institute of General Medical Sciences of the National Institutes of Health (NIH) (T32-GM007280; R35-GM124836; R35-GM138113); National Institute of Diabetes and Digestive and Kidney Diseases (U24-DK062429); National Human Genome Research Institute of the NIH (R01-HG010365).
