Machine Learning Predicts Risk of Aneurysm
News Sep 19, 2018 | Story sourced from the National Institutes of Health
An aneurysm can develop and grow for years without causing any symptoms. They’re often found by chance when a patient is getting images—like MRI or CT scans—done for another reason. Photo credit: Pixabay
An abnormal bulge in a blood vessel, known as an aneurysm, usually has no symptoms, but can be deadly if it expands and bursts. An aneurysm that occurs in the main artery that leads from the heart through the belly is called an abdominal aortic aneurysm (AAA). Some people are lucky enough to have AAA detected during medical scans for other reasons.
Being over 65, being male, smoking, having high blood pressure, and having a buildup of plaque in the arteries are AAA risk factors. A family history of AAA is also thought to play a role. Lifestyle changes and treatments may prevent an aneurysm from expanding and bursting.
A research team led by Drs. Philip S. Tsao and Michael Snyder of Stanford University set out to develop a way to predict which people are at risk of having AAA. They used genome sequences and machine learning techniques to create an algorithm they call HEAL (Hierarchical Estimate from Agnostic Learning). The work was funded in part by NIH’s National Heart, Lung, and Blood Institute (NHLBI) and National Human Genome Research Institute (NHGRI). Results were published in Cell on September 6, 2018.
The scientists performed whole genome sequencing on blood samples from 133 healthy people and 268 people known to have AAA. In the people with AAA, medical scans showed the artery had ballooned from a normal diameter of about 2 centimeters to at least 3 centimeters.
Genome sequencing identified nearly 24 million genetic mutations. Of these, the scientists considered about 66,000 rare mutations that weren’t found in previous searches for common mutations in healthy people. The machine learning sytem analyzed this data and identified 60 genes that were more likely to have elevated mutations in the people with AAA.
When the team tested HEAL on their sample, it could correctly distinguish which people had AAA based on their genomes 69% of the time. When smoking history, cholesterol levels, and other data from surveys and health records were included, HEAL was able to distinguish AAA 80% of the time.
HEAL also identified biological processes that may be involved in development of AAA. These include the immune response and blood vessel development.
“What’s important to note about AAA is that it’s irreversible, so once your aorta starts enlarging, it’s not like you can un-enlarge it. And typically, the disease is discovered when the aorta bursts, and by that time it’s 90% lethal,” Snyder explains. “No one has ever set up a predictive test for it and, just from a genome sequence, we found that we could actually predict with about 70% accuracy who is at high risk for AAA.”
With further development, this work could lead to ways to identify people who are at risk for AAA. The results also suggest new research directions and potential therapeutic targets. In addition, this study is a proof of the principle that machine learning can be used to predict disease risk for other genetic conditions.
This article has been republished from materials provided by the National Institutes of Health. Note: material may have been edited for length and content. For further information, please contact the cited source.
Li, J., Pan, C., Zhang, S., Spin, J. M., & Snyder, M. (2018). Decoding the Genomics of Abdominal Aortic Aneurysm. Cell. doi:10.1016/j.cell.2018.07.021.
Researchers have created a portable and fast-acting test that can distinguish Ebola infections from other fever-causing infectious diseases such as Lassa fever and malaria in around 30 minutes. Although further testing is required, this could be useful during febrile disease outbreaks.READ MORE