Using genomics and machine learning to predict aneurysm risk

DNA genomics precision medicine
Stanford University researchers developed an algorithm they believe can be used to identify the genetic underpinnings of several diseases. (Pixabay)

Abdominal aortic aneurysm (AAA), an irreversible disorder in which the body’s main artery gradually enlarges, affects more than 3 million Americans per year. If the artery bursts, it’s almost always fatal, so doctors who treat AAA have long been interested in being able to identify people who face a higher-than-average risk of developing the condition.

Scientists at Stanford University have developed a way to forecast AAA using genomics and machine learning. Rather than attempting to tie the disorder to individual genetic mutations—like testing for BRCA1 or BRCA2 mutations to predict breast cancer risk—they analyzed genomic data from 268 people who had AAA to look for patterns of mutations.

The team identified 60 genes that were hypermutated in AAA patients, according to the study, published in the journal Cell. Some of the genes were previously known to be important in blood-vessel development, but others were more closely associated with immunity. To validate their findings, the researchers scanned the genomes of 133 healthy people and discovered that they did not have the same abnormal mutational patterns.

Your Daily Newsletter — Free

Enjoying this story? Subscribe to FierceBiotech!

Biopharma is a fast-growing world where big ideas come along every day. Our subscribers rely on FierceBiotech as their must-read source for the latest news, analysis and data in the world of biotech and pharma R&D. To read on the go, sign up today to get biotech news and updates delivered right to your inbox!

RELATED: FDA panel backs Cordis’ Incraft stent graft for abdominal aortic aneurysms

The methodology the Stanford team used was “agnostic,” said Michael Snyder, Ph.D., professor and chair of genetics at the university, in a statement. That means they didn’t go into the study suspecting that particular genes might play a role in AAA. Instead, they fed genetic information from patients into an algorithm and “let machine learning figure it out,” he said.

The algorithm and genomic sequencing alone predicted AAA risk with about 70% accuracy, the researchers reported. When they added information from electronic medical records, such as smoking habits or high blood pressure history—both of which increase the risk—accuracy improved to 80%. Such a test could be useful to physicians, they say, because once AAA sets in, it can’t be reversed, and it’s often not discovered until the aorta bursts. Lifestyle changes, such as increasing levels of HDL (the “good” cholesterol), can delay the onset of the disease.

Using artificial intelligence to improve the diagnosis and prevention of serious diseases is an area of intense research. In May, the Mayo Clinic and AliveCor reported they used AI to identify people who face a heightened risk of arrhythmias and sudden cardiac death in spite of being diagnosed as healthy with a standard electrocardiogram. WuXi NextCODE Health is developing products designed to help consumers, physicians and drug companies use genomic data to predict disease risk.

Stanford’s Snyder envisions a future in which every person undergoes genomic sequencing shortly after birth. “Both your single-gene and your complex disease risk will be used to predict your overall disease risk, and then you can take action based on that information,” he said. His team’s next step is to adapt their AAA-predicting algorithm to uncover the genetic causes of preterm birth and autism.