Machine Learning May Benefit Future Alzheimer’s Patients

IBM researchers develop a possible blood test for early identification of Alzheimer’s disease.

Alzheimer’s disease has challenged researchers for decades. It is known that the protein amyloid-beta is produced at higher levels in Alzheimer’s patients and may be associated with the formation of deposits in the brain that are indicative of the disease. It is also known that this protein is present in the spinal fluid of patients well before memory loss begins to occur and up to 10 years before PET imaging reveals signs of Alzheimer’s. Detecting it, however, currently requires invasive surgery performed under anesthesia, preventing its use as an early detection method.

IBM researchers in Australia may have a noninvasive solution. They have developed a blood test that predicts the buildup of amyloid-beta in spinal fluid with up to 77% accuracy.

The test was developed using machine learning and data downloaded from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), a long-term imaging study of older adults examining whether MRI and PET images can be combined with other measures to predict early-stage Alzheimer’s. Images from 566 people were used in their study, 182 of whom were carriers of APOEε4, a gene variant that raises the risk of late-onset Alzheimer’s.

This data was combined with information on nearly 400 proteins that were measured in blood samples from the ADNI participants. To develop a simplified test, the researchers used machine learning to identify the minimal number of proteins needed to reasonably reflect the buildup of amyloid-beta. Four proteins that can be detected in blood samples –– chromogranin-A (CGA), Aβ1–42 (AB42), eotaxin 3, and apolipoprotein E (APOE) –– were found to be the best predictors.

The model was validated using a predictive model based on age and APOEε4 status and through comparison to blood tests using different variables. All of the models using blood analytes outperformed the model using age and APOEε4, but the models that used the protein level measurements consistently provided the strongest predictive ability.

The four-protein blood test could potentially be used to select patients for clinical trials of Alzheimer’s drugs and for early detection and treatment of the disease. Several recent clinical trials for new treatments of Alzheimer’s disease have failed recently, possibly because the patients involved in the studies were in the latest stages of the disease and not likely to respond well to treatment.


Emilie Branch

Emilie is responsible for strategic content development based on scientific areas of specialty for Nice Insight research articles and for assisting client content development across a range of industry channels. Prior to joining Nice Insight, Emilie worked at a strategy-based consulting firm focused on consumer ethnographic research. She also has experience as a contributing editor, and has worked as a freelance writer for a host of news and trends-related publications