Harnessing AI and Laboratory Tests to Assess Disease Risks from Rare Genetic Variants

Researchers at the Icahn School of Medicine at Mount Sinai have developed an innovative approach that combines artificial intelligence (AI) with routine laboratory tests to better understand the health risks associated with rare genetic mutations. Often, when genetic testing uncovers uncommon DNA changes, both doctors and patients may struggle to interpret their significance. This new method aims to provide a more precise prediction of disease development—a concept known as penetrance.

The team utilized AI algorithms trained on extensive electronic health records (EHRs), along with standard lab data such as cholesterol levels, blood counts, and kidney function tests. Their study, published in Science under the title "Machine learning-based penetrance of genetic variants," leverages these data to quantify the likelihood that an individual with a specific genetic variant will develop a related disease.

Traditional genetic assessments tend to classify results in binary terms—either the presence or absence of a disease. However, many conditions like high blood pressure, diabetes, or certain cancers exist on a spectrum, and their risk factors are complex. To address this, the AI models were designed to evaluate disease risk on a scale, offering more nuanced insights into how genetic variants might influence health.

"Our goal was to move past black-and-white answers that leave uncertainty," explained Ron Do, Ph.D., senior author and professor at Mount Sinai. "By integrating AI and real-world lab data, such as cholesterol and blood counts, we can more accurately estimate disease risk for individuals with specific genetic changes. This scalable approach enhances precision medicine, especially when dealing with uncertain or rare findings."

Using data from over one million EHRs, the researchers crafted models for ten common diseases. These models were then applied to individuals known to possess rare genetic variants, generating a risk score between 0 and 1. A higher score indicates a greater likelihood that the variant contributes to disease, while a lower score suggests minimal risk. They analyzed more than 1,600 genetic variants, some of which unexpectedly showed significant disease signals previously deemed uncertain, and others thought to be harmful that exhibited little real-world impact.

The findings highlight that AI-driven scores can assist clinicians in making more informed decisions. For example, a high-risk score for a variant linked to Lynch syndrome could prompt earlier cancer screening, while a low score might prevent unnecessary interventions. Though not intended to replace clinical judgment, these scores serve as valuable guides in managing genetic test outcomes.

The research team aims to further refine their models by including more diseases, genetic variations, and diverse populations. They are also interested in tracking real-world outcomes over time to assess the predictive accuracy and potential benefits of early preventive measures.

"This represents a promising step toward personalized medicine, where AI and routine clinical data work together to produce actionable insights," noted Dr. Do. "Our vision is to make genetic risk assessment more accessible, understandable, and ultimately supportive of better patient care."

Source: https://medicalxpress.com/news/2025-08-ai-lab-combine-disease-rare.html