Measuring proteins that circulate in the blood could boost the performance of standard diagnostic approaches for dozens of medical conditions, supporting the earlier detection and treatment of conditions such as cancer, heart disease and motor neuron disease.
Researchers have previously used blood protein levels to identify people at high risk of a narrow range of common conditions, such as and . But it is unclear whether this approach boosts the performance of existing clinical models that rely on basic information 鈥� such as someone鈥檚 age, weight to height ratio and smoking status 鈥� along with data from common blood tests, which look at blood cells and only around a dozen proteins.
To find out more, at the University of Cambridge and her colleagues turned to health records and data that contained information on roughly 3000 proteins found in blood, previously collected from more than 41,000 people.
Advertisement
Using information from around 70 per cent of these participants, the team built statistical models that predict the risk of developing a range of both rare and common conditions within a 10-year follow-up period.
By testing the models on the remaining 30 per cent of the participants, the researchers found that combining the protein-based model with the standard clinical models allowed them to pinpoint the risk of developing 52 medical conditions.
The team identified around 70 per cent of the participants who went on to develop dilated cardiomyopathy (where the heart鈥檚 muscle walls become stretched and thin) and pulmonary fibrosis (scarring of lung tissue). In contrast, the clinical models alone identified roughly 30 per cent of those who went on to get these conditions.
The combined approach also identified more than twice as many participants who went on to develop bone marrow cancer and motor neuron disease, compared with the clinical models alone.
Combining the protein-based and clinical models incorrectly classed people as being at high risk for a particular condition 10 per cent of the time. The acceptable rate for such 鈥渇alse positives鈥� will vary depending on how common the condition is, the value of detecting the risk early and the cost of uncovering false positives and carrying out a correct diagnosis, says at SomaLogic in Colorado, a company that uses artificial intelligence and proteins to predict a person鈥檚 medical condition risk.
If applied in clinics, people identified as being at high risk for certain conditions could be screened more frequently, allowing earlier detection and treatment, according to the team.
鈥淭his is further evidence that protein biomarkers in the blood offer great promise for improved and earlier diagnosis, across a much wider range of diseases,鈥� and at the University of Liverpool, UK, told 快猫短视频 in a joint statement.
However, the participants were mainly of European ancestry and living in the UK, so the findings must be validated in studies of more ethnically diverse populations and across more locations, says Williams.
medRxiv
