AI Model Uses Lab Tests to Predict Genetic Disease Risk

Traditional studies usually classify conditions as present or absent, yet many diseases, such as diabetes, hypertension, and cancer, develop along a spectrum. This lack of clarity has long limited the ability to act on ambiguous genetic findings. Now, researchers have developed a powerful new way to determine whether a patient with a mutation is likely to actually develop the disease.

Researchers at the Icahn School of Medicine at Mount Sinai (New York City, NY, USA) have developed an artificial intelligence (AI) tool that combines machine learning with routine lab data and electronic health records. Reported in Science, the method integrates common lab values such as cholesterol, blood counts, and kidney function with AI-driven models. This approach moves beyond black-and-white risk assessments and provides a scalable, data-rich view of genetic penetrance, or the likelihood that a mutation leads to disease.

Using over one million electronic health records, the research team built AI models for 10 common diseases. These models were then applied to individuals carrying rare variants to generate a “machine learning penetrance” score ranging from 0 to 1, where a higher score suggests greater disease risk. The researchers calculated penetrance estimates for more than 1,600 variants and found that some mutations previously labeled as uncertain showed strong disease associations, while others assumed to be harmful appeared benign.

The findings reveal that AI models can provide more clinically relevant insights than binary categorizations. In one example, a patient with a high-scoring variant linked to Lynch syndrome might benefit from earlier cancer screening, whereas a low-risk score could help avoid unnecessary interventions. While not intended to replace clinical judgment, the AI model can guide physicians toward more tailored preventive strategies.

The team plans to expand the model to cover more diseases, a broader range of genetic mutations, and more diverse populations. Future studies will evaluate whether individuals flagged as high risk by the AI system actually go on to develop disease, and whether early interventions can alter outcomes. These steps are key to validating the model’s utility in precision medicine and real-world care.

“We wanted to move beyond black-and-white answers that often leave patients and providers uncertain about what a genetic test result actually means,” said Ron Do, PhD, senior study author and the Charles Bronfman Professor in Personalized Medicine at the Icahn School of Medicine at Mount Sinai. “By using artificial intelligence and real-world lab data, such as cholesterol levels or blood counts, we can now better estimate how likely disease will develop in an individual with a specific genetic variant.”

Related Links:
Icahn School of Medicine at Mount Sinai