When genetic testing reveals rare DNA mutations, doctors and patients are often left in the dark about what it actually means. Now, researchers at ICAHN School of Medicine in Mount Sinai have developed a powerful new method to determine whether a mutant patient is more likely to actually develop a disease, a concept known as permeability.
The team set out to solve this problem using artificial intelligence (AI) and routine lab tests such as cholesterol, blood count, and kidney function. Details of the survey results were reported in the online issue on August 28th Science. Their new methods combine machine learning with electronic health records to provide more accurate, data-driven genetic risks.
Traditional genetic studies often rely on simple YES/NO diagnoses to classify patients. However, many diseases such as hypertension, diabetes and cancer do not fit properly into binary categories. Researchers at Mount Sinai trained AI models to quantify diseases on the spectrum and provide more nuanced insight into how disease risks occur in real life.
“We wanted to move beyond the black and white answers that often leave patients and providers uncertain about what genetic test results actually mean,” says Rondeau, a personalized medicine professor at Icahn School of Medicine in Mount Sinai. “By using artificial intelligence and real-world lab data such as cholesterol levels and blood counts, which are already part of most medical records, we can better estimate how subtle, scalable, scalable and accessible the disease is in an individual with a particular genetic variant.
Using more than one million electronic health records, researchers have built AI models for 10 common diseases. We then applied these models to people known to have rare genetic variants, generating scores between 0 and 1, reflecting the likelihood of developing the disease.
A high score close to 1 suggests that the variant is more likely to contribute to the disease, while a low score indicates minimal or no risk. The team calculated “ML penetration” scores for over 1,600 genetic variants.
Some of the results were surprising, investigators say. Variants previously labelled “uncertain” showed clear disease signals, while other variations were thought to cause disease, but had little effect on actual data.
“Our AI models are not intended to replace clinical judgment, but they may serve as potentially important guides, especially when test results are unknown. Physicians can use ML penetration scores in the future to avoid unnecessary worries and interventions, in order to avoid whether patients undergo previous screening, take precautions, or whether variations are at low risk, Sinai. “For example, if a patient has a rare variant associated with Lynch syndrome, it may score a high score and cause previous cancer screening, but if the risk is low, a jump to conclusion or overtreatment may be avoided.”
The team is currently working to expand the model to include more diseases, broader genetic changes and more diverse populations. They also plan to track how well these predictions endure over time, whether people with high-risk variants actually develop the disease, and whether early behavior can make a difference.
“Ultimately, our research points to a potential future in which AI and routine clinical data work closely together to provide personalized, actionable insights by patients and families navigating genetic test results.
The paper is titled “Machine Learning-Based Penetration of Genetic Variants.”
The authors of the studies listed in the journal include Iain S. Forrest, Ha My T. Vy, Ghislain Rocheleau, Daniel M. Jordan, Ben O. Petrazzini, Girish N. Nadkarni, Judy H. Cho, Mythily Ganapathi, Kuan-Lin Huang, Wendy K. Chung, and Ron Do.
This work was supported in part by the following grants: NIH National Institute of Medicine (R35-GM124836); National Institute of Diabetes and Gastrointestinal Diseases and Kidney Diseases (U24-DK062429); NIH National Institute of Human Genome (R01-HG010365); NIH National Institute of Medicine (R35-GM138113); National Institute of Diabetes and Gastrointestinal Diseases and Kidney Diseases (U24-DK062429).
*Mount Sinai Health System Member Hospital:Mount Sinai Hospital. Mount Sinai Brooklyn; Mount Sinai Morning Side. Mount Sinai Queens; Mount Sinai South Nassau. Mount Sinai West; and New York Eye and Ear Treatment on Mount Sinai
