What is the risk of a person developing dementia? A new study suggests that relying on blood tests may provide predictive results 15 years in advance. Recently, a team led by Professor Yu Jintai from the Department of Neurology at Huashan Hospital affiliated with Fudan University, Professor Feng Jianfeng and Researcher Cheng Wei from the Institute of Brain Intelligence Science and Technology at Fudan University collaborated with scientists from the University of Warwick in the UK to use AI algorithms to analyze 1463 blood proteins, identify several biomarkers, and predict a person's risk of developing dementia 15 years in advance. The relevant paper was published in the journal Nature Aging. Yu Jintai and Feng Jianfeng emphasized in an interview with Science and Technology Daily that the latest research results are of great significance for screening and early intervention of high-risk populations for dementia. This study is expected to lead to new blood testing methods that can detect Alzheimer's disease and other forms of dementia as early as possible. In addition, the strong collaboration between AI and protein analysis is also of great significance in promoting the development of precision medicine. According to data from the World Health Organization, over 55 million people worldwide suffer from dementia each year, relying on machine learning to identify biomarkers. This disease is very cunning, and patients may have been diagnosed with it for many years. "The course of dementia is long, and pathological changes may occur several years or even decades before onset. By the time of diagnosis, the optimal treatment window is often missed. Therefore, it is urgent to find early prediction methods for dementia and identify high-risk individuals for dementia," explained Yu Jintai. In order to identify proteins that can reveal clues to dementia, the team used the UK Biobank database to collect and freeze blood samples from 52645 healthy participants between 2006 and 2010. During a follow-up period of approximately 15 years, 1417 of these individuals developed dementia. Feng Jianfeng introduced: "In the blood of dementia patients, the levels of certain specific proteins can change. We used machine learning methods to analyze 1463 proteins and found that the elevated levels of GFAP, NEFL, and GDF15 are related to dementia. We used machine learning algorithms to combine the above biomarkers with demographic factors such as age, gender, education, and family history of the subjects.". The results show that under different validation strategies, the accuracy of the model in predicting the incidence rate of three dementia subtypes, including Alzheimer's disease, is about 90%. " GFAP can provide structural support for astrocytes and has previously been proposed by scientists as a biomarker for diagnosing Alzheimer's disease. Yu Jintai added that the latest research has found that people with high levels of GFAP in their blood are 2.91 times more likely to develop Alzheimer's disease than those with normal levels. "Compared to the previous diagnosis time, we can reliably predict the risk of this type of disease 15 years in advance," said Feng Jianfeng. Yu Jintai believes that this discovery can be used to develop new blood testing methods to identify individuals at risk of dementia as early as possible. By conducting a simple blood test to check for the presence of disease marker proteins in the blood, this method is relatively more invasive or cost-effective compared to other detection methods such as lumbar puncture, imaging examination, etc