Chinese scientists use massive databanks and AI to try to predict dementia 15 years before symptoms start

Using a data-driven strategy, scientists “innovatively identified important plasma biomarkers for predicting future dementia,” the team reported in a paper published Monday in the peer-reviewed journal Nature Aging. I wrote

But scientists have bigger goals in mind for blood biomarker tools, such as using them to accurately predict whether a patient might develop the disease in the future, even before they show any clinical symptoms. before

According to the authors of the paper, there is no cure for dementia and it is worth understanding that if a person can develop it, it can allow for early diagnosis and intervention.

The large-scale study of proteins – also known as proteomics – can be used to find potential drug or diagnostic interventions for diseases and to better understand how the human body works.

However, the team wrote that systematically studying the protein in blood proved difficult due to “technical constraints” and a lack of comparative methods.

To overcome this hurdle, the team drew heavily on the UK Biobank cohort, which enrolled more than 50,000 people aged 40 to 69 and followed a median of 14 years starting in the mid-2000s. The follow-up period was

Just over 1,400 subjects in the biobank cohort – who provided all biological samples and demographic information – developed dementia within 10 years of initial data collection.

The biobank recently released a new data set of 1,400 plasma, or blood, proteins found in participants’ samples during initial intake and follow-up sessions.

Decoding Dementia: The SCMP Lifestyle Series presents research on causes and treatments.

This data release gave the team an “unprecedented opportunity” to conduct a proteomics study on blood proteins associated with the development of dementia.

This allowed them to “track plasma protein trajectories from the time of dementia diagnosis and estimate when each protein begins to deviate from normal control values,” they wrote.

The scientists found hundreds of associated proteins, but they focused their studies on a handful of “key proteins” that had begun to change expression at least a decade before the clinical onset of dementia.

Machine learning algorithms were used to determine which proteins formed a better predictive model and checked this against biobank data that identified which subjects developed dementia.

Yu said the algorithm, which played a “crucial” role in the research, had “powerful pattern recognition and prediction capabilities,” allowing for more efficient screening of large-scale data sets.

According to the team, incorporating protein data by itself into a predictive model was “unlikely to achieve the highest predictive accuracy.”

To develop an “optimal predictive algorithm that is non-invasive, cost-effective and easily accessible”, they combined data on a protein called GFAP – which they found more than doubled the risk of dementia. – associated with demographic information such as age. and sex.

Their final, combined predictive model shows promise in being able to provide “accurate predictions of future dementia, up to 10 years before diagnosis,” according to the paper.

And compared to imaging scans or spinal taps used to screen people for disease risk, their method could also “offer substantial cost advantages,” the paper said.

Beijing has emphasized advances in chips and quantum computing to command the future.

The team said their study has some limitations, with more than 90 percent of the biobank cohorts consisting of subjects of white ethnicity and therefore not representative of the world. The proteins examined also do not cover the entire human proteome.

However, Yu said the team is now also researching a cohort of Chinese, which will allow them to examine similarities and differences with the UK-based database.

His work so far has “provided clues for the development of new treatment and intervention strategies,” he said, adding that the team is looking at other brain-related conditions such as depression and Parkinson’s disease. Using biobank.