Can you imagine knowing how high the risk of developing dementia is based on just a single drop of blood testing assay? A protein test of blood plasma can predict the risk of dementia 15 years in advance. In the near future, people will be able to know in advance how likely they are to develop dementia from a blood test report card.

Recently, a team of researchers from Fudan University used large-scale proteomics data and artificial intelligence algorithms to discover an important plasma biomarker for predicting the risk of dementia in the future, and the related research results were published in Nature Aging. The main issue of Nature commented that the study “marks a step forward towards blood tests that can detect Alzheimer’s disease and other types of dementia in the early asymptomatic stages.”

It is worth mentioning that this study used AI for science (Note: Artificial Intelligence-driven Scientific Research, hereinafter referred to as AI4S) to analyze and model 1,463 types of plasma proteomics data, and mined out key biomarkers that are predictive of the risk of developing dementia, which opens up the possibility of early intervention and early treatment of the disease.

How can AI4S help early monitoring of brain diseases?

Dementia patients, represented by Alzheimer’s disease, and their families are under tremendous pressure psychologically, physically and financially. Researchers point out that there is an insidious period of years or even decades before patients’ clinical symptoms appear. Early manifestations are often confused with normal aging, and by the time a patient develops significant cognitive behavioral deficits and goes to the hospital, the disease has often progressed to the mid-to-late stages, missing the best time for intervention.

There is currently no effective treatment for dementia, and actively promoting the early identification of dementia so as to achieve early intervention and treatment is the key to reducing the burden of the disease.2023 In April 2023, Prof. Jintai Yu’s team published a research paper in Nature Human Behavior, a sub-journal of Nature, stating that lifestyle, medical history, and socio-economic status are associated with the majority of dementia, with an estimate of as high as 47.0% -72.6% of dementia cases are preventable, and good living conditions, lifestyle, physical indicators, reducing co-morbidities and improving socio-economic status can mitigate the risk of dementia from genetics.

Based on AIS4S, using the world’s largest community cohort-based proteomics data and AI algorithms to date, the team of Prof. Jianfeng Feng/Researcher Wei Cheng at the Institute of Brain-Like Intelligence Science and Technology of Fudan University, together with Prof. Jintai Yu’s team at Huashan Hospital of Fudan University, have launched a joint research project.

Feng explained that unlike previous similar studies that used a small sample size cross-sectional design, this study utilizes a large sample of longitudinal data over a long period of time to extract useful patterns, trends, and correlational information, and to use the data to “speak for itself”. “Our study provides a good research example of AIS4S. Based on the data-driven idea, we constructed a high-precision dementia risk prediction model, which is a breakthrough progress in the cross-fertilization of science, technology and medicine, and is of great significance in promoting the development of precision medicine.” Feng said.

Based on the large sample cohort data, the team followed up 52,645 non-demented community people for an average of over 14 years, of which 1,417 participants were diagnosed with new-onset all-cause dementia (ACD), 691 patients with new-onset Alzheimer’s disease (AD), and 285 patients with new-onset vascular dementia (VaD). By analyzing data on 1,463 plasma proteins, the team identified plasma biomarkers that are highly valuable for dementia prediction.

After modeling and machine learning algorithm analysis, three plasma proteins, GFAP, NEFL, and GDF15, were consistently associated most significantly with the risk of new-onset ACD, AD, and VaD, the team said. Analysis of the associations between different plasma protein levels and the risk of clinical progression of the disease revealed that subjects with higher baseline GFAP, NEFL, or GDF15 levels had a substantially increased risk of developing dementia in the future. For example, those with higher baseline GFAP levels were 2.32 times more likely to develop dementia in the future than those with lower baseline GFAP levels.

Inexpensive, convenient and non-invasive
Accuracy improved to 90 percent

“Detecting brain disease is often difficult. Because lumbar puncture is invasive, imaging is expensive, and clinical implementation of related technologies is limited by space and other constraints, making it difficult to generalize. Hematology testing is convenient, non-invasive and inexpensive, and can serve as an ideal tool for early risk screening of a wide range of people at the preclinical stage.” Cheng Wei explained.

Yu Jintai said, “This discovery of important plasma biomarkers provides a new theoretical basis for the transition of hematology testing from research to clinical. Moreover, the blood test indicators we discovered this time are more simple, accessible and easy to popularize, and can do a good job of predicting both the short-term risk of dementia onset and the risk of dementia onset more than ten years later.”

According to the report, this study can predict the risk of dementia onset 15 years in advance and with an accuracy exceeding 90%. “This shows that proteomics can play an important role in the early precision identification and intervention of brain diseases, and provides new ideas for future brain disease research.” Cheng Wei said, through the blood test, it is expected to assist clinicians to identify high-risk patients with dementia as early as possible, and intervene as early as possible to improve patients’ quality of life.

Half a year later, it is expected to be used for general public testing

The application of this discovery, how far away from the general public? According to reports, the researchers are in commercialization talks with the company, the goal is to make this test, which currently costs up to hundreds of pounds, more accessible, and is expected to be applied to clinical testing in half a year to screen out high-risk groups. This opens up the possibility of early intervention to slow or even eliminate the progression of hindering conditions.

The research team revealed that some medical check-up medical institutions have taken the initiative to contact the team to explore the possibility of adding the relevant test to their medical check-up programs. The next step will be to conduct data collection and cross-validation around China’s dementia risk population cohort, make corrections to the relevant data, and develop a dementia risk prediction data model that is most suitable for China’s population.

This interdisciplinary team of top brain science, artificial intelligence, and neurology experts at Fudan University was formed in 2021 and has jointly published more than a dozen topical articles in related fields. The full-phenotype dementia prediction model they constructed earlier has managed to predict the risk of onset of the disease 10 years in advance, with an accuracy of 85%, and this study has advanced the prediction period to 15 years before onset, with a prediction accuracy exceeding 90%.

According to relevant data, more than 55 million people worldwide are currently living with dementia, and this number is expected to reach 78 million by 2030. About 70 percent of all dementias are caused by Alzheimer’s disease, with vascular dementia caused by vascular damage accounting for 20 percent of cases.

Translated with www.DeepL.com/Translator (free version)