[ad_1]
The researchers used data from the British biobank – both a long-term study and a genetic repository including genomic data on about half a million people – to search for risk factors for varicose veins to help of machine learning combined with epidemiological methods. In addition, they searched for genetic markers using genome association studies in 337,536 participants, of whom 9,577 had varicose vein disease. The study confirmed that currently established risk factors – including age, woman, overweight or pregnancy, or a history of deep vein thrombosis – are all associated with varicose veins.
"We have confirmed that deep vein thrombosis in the past exposes you to increased risk in the future," Leeper said. "Recent research suggests that the opposite also seems to be true. Having varicose veins exposes you to these clots.
The study also confirmed that leg surgery, family history, lack of movement, smoking and hormone therapy are risk factors. But the correlation they found between size and condition was unexpected, according to the researchers.
"We were very surprised to find that our machine learning analyzes were up to par," said Flores.
Disable the algorithm
Typically, in a large-scale genetic study like this, researchers use genome-wide association studies to examine DNA variations that may be associated with increased risk for the genome. a particular disease. Using this method, researchers identified 30 regions of the genome associated with varicose veins. But researchers have also used another method involving machine learning, a type of artificial intelligence, to launch a giant net to uncover any previously unknown risk factors.
"These methods represent new ways of thinking about research," Ingelsson said. "You enter without hypothesis on a specific biological mechanism and look for something new. You could say that you let go of the machine on it. In this case, we included 2,716 predictors of varicose veins in this machine learning algorithm. Then we let the algorithms find the strongest predictors of varicose veins. "
In addition to size, the machine learning algorithm has shown that bioimpedance, a measure of the body's ability to impede the flow of electrical current, is a powerful predictor of varicose veins. This measure could potentially be used as a diagnostic tool to predict varicose veins, Leeper said.
When the machine learning analysis revealed that height was a possible risk factor, the researchers conducted additional tests to determine if it was a real cause of the disease with the help of Mendelian analyzes, a statistical technique for determining causal effects.
"Our results strongly suggest that size is a cause, not just a correlated factor, but an underlying mechanism leading to varicose veins," Ingelsson said.
He added, "By doing the largest genetic study ever done on varicose veins, we now have a much better understanding of modified biology in people at risk."
Daniela Zanetti, PhD, a postdoctoral researcher at Stanford, also contributed to the study, as did researchers at the University of Uppsala in Sweden.
The study was funded by the National Institutes of Health (grant 1R01HL135313) and the Knut Foundation and Alice Wallenberg.
The Stanford Department of Medicine also supported the work.
Source link
