Machine learning models, blood biomarkers in research predict future disease risk

Artificial Intelligence Ai Suit Social

Applying predictive machine-learning models to genetic data and blood biomarkers can provide information about the risk of developing disease up to ten years into the future, according to research shared at the European Society of Human Genetics.

There is rising interest in using genomic data to generate polygenic risk scores that predict the likelihood of an individual developing certain diseases. The new research found that analyzing genomic information and blood biomarkers is more accurate and cost-effective than analyzing genomic data alone, suggesting the combined approach can play a role in helping healthcare systems become more proactive.

Jeffrey Barrett, the chief scientific officer at Nightingale Health in Helsinki, and colleagues carried out the research into the effectiveness of combined analyses of biomarkers and genomic information. The team measured more than 200 biomarkers in blood samples from 300,000 participants in the UK Biobank and 200,000 people in the Estonian biobank.

“This is the biggest dataset of its kind that has ever been collected,” Barrett said in a statement. Using machine learning, the researchers built models to predict individuals’ future risks of developing certain diseases based on their genetic information and biomarkers. The diseases were ischemic heart disease, stroke, lung cancer, diabetes, chronic obstructive pulmonary disease, Alzheimer’s and other dementias, depression, liver disease, and colon cancer.

“We found that in all the diseases, both genetics and biomarkers could provide useful information about disease risk, even ten years into the future. The blood biomarkers provided better prediction in nearly all cases,” Barrett said.

The 10% of people with the highest risk of lung cancer based on biomarkers had four times the risk of the average person. The top 10% of people based on genetics had 1.8 times greater risk than average. For liver disease, the increased risk based on biomarkers was 10 times, versus two times for genetics.

For some diseases, blood biomarkers were particularly predictive of risk in the next two to four years. The finding may reflect links between what the biomarkers measure and the presymptomatic phase of the disease. If validated, the finding suggests healthcare systems may be able to use blood tests to find patients who would benefit from preventative health actions.

“It means that it is relatively easy to find the individuals at greatest risk of many diseases and offer them ways to reduce their risk, keeping them healthier and at the same time reducing the financial burden on healthcare systems,” Barrett said.

Page 1 of 194
Next Page