2024 Alzheimer's Association Research Grant (AARG)
Novel AD Plasma Biomarker Machine Learning Analysis & Molecular Validation
Can novel computer science analyses help identify the wide range of molecules in the blood that may promote Alzheimer’s risk?
Amy Tsurumi, Ph.D.
Massachusetts General Hospital
Boston, MA - United States
Background
Because Alzheimer’s gets worse over time, there is a need to find improved ways of diagnosing the disease at an early stage, when treatments can be most effective. Current diagnostic methods are often cumbersome and expensive. They can require invasive imaging procedures and lumbar punctures (spinal taps) in order to identify biomarkers (biological factors) that indicate disease. Researchers, therefore, are exploring new diagnostic methods that are safer, less expensive, and can offer relatively quick results. One such method might involve blood tests. Studies show that dementia-related protein levels in the blood (such as levels of beta-amyloid and tau protein) may predict similar changes in the brain. In addition, blood tests require only a relatively simple drawing of blood from the individual being diagnosed. Blood-based biomarkers for dementia, however, are numerous and complex, and many have not yet been discovered. A more complete understanding of these biomarkers will be necessary to make a future Alzheimer’s blood test truly viable.
In preliminary studies, Dr. Amy Tsurumi and colleagues have been searching for novel dementia-related biomarkers in the blood with advanced computer science techniques called machine learning. They found that by analyzing blood samples with multiple machine learning methods, they could develop a more complete array (or panel) of such biomarkers.
Research Plan
The researchers will conduct a larger study to test and verify their machine-learning method. First, they will analyze blood samples from a large collection at their hospital called the Massachusetts General Brigham (MGB) Biobank, as well as blood data from a study of aging called the Alzheimer’s Disease Neuroimaging Initiative (ADNI). These samples and data are from people with late-onset Alzheimer’s, the most common form of the disease. Dr. Tsurumi’s team will use the results of this analysis to identify a model of blood-based protein and messenger RNA (mRNA) biomarkers that best predict Alzheimer’s. Messenger RNA is the genetic material cells use to make proteins from genes. Next, the researchers will collect clinical electronic health records from more than 14,000 individuals with Alzheimer’s who attended their hospital. They will then use machine learning analyses to develop a model for predicting Alzheimer’s that includes a wide variety of clinical factors (including age, sex, ethnicity, blood pressure, and brain cell changes). Finally, the team will compare this model with their protein/mRNA model to develop a more thorough, combined model for predicting dementia.
Impact
If successful, Dr. Tsurumi’s combined model could be tested more extensively in future studies – studies with data from a more diverse array of participants. Ultimately, this work could lead to an effective, targeted blood test for Alzheimer’s.