Signatures built from the vast array of proteins found in the blood stream, including those encapsulated in extracellular vesicles, should in principle correlate with many health conditions. This includes those conditions, such as Alzheimer's disease, characterized by a long, slow preclinical stage in which damage and metabolic disarray builds up over time. Modern machine learning techniques allow the cost-effective construction of such signatures, given a large enough database work with, and as illustrated here.
Efforts to gauge people's Alzheimer's risk before dementia arises have focused mainly on the two most obvious features of Alzheimer's brain pathology: clumps of amyloid beta protein known as plaques, and tangles of tau protein. Scientists have shown that brain imaging of plaques, and blood or cerebrospinal fluid levels of amyloid beta or tau, have some value in predicting Alzheimer's years in advance. But humans have tens of thousands of other distinct proteins in their cells and blood, and techniques for measuring many of these from a single, small blood sample have advanced in recent years. Would a more comprehensive analysis using such techniques reveal other harbingers of Alzheimer's?
An initial analysis covered blood samples taken during 2011-13 from more than 4,800 late-middle-aged participants in the Atherosclerosis Risk in Communities (ARIC) study, a large epidemiological study of heart disease-related risk factors and outcomes, recording levels of nearly 5,000 distinct proteins in the banked ARIC samples. The researchers analyzed the results and found 38 proteins whose abnormal levels were significantly associated with a higher risk of developing Alzheimer's in the five years following the blood draw.
Researchers then measured protein levels from more than 11,000 blood samples taken from much younger ARIC participants in 1993-95. They found that abnormal levels of 16 of the 38 previously identified proteins were associated with the development of Alzheimer's in the nearly two decades between that blood draw and a follow-up clinical evaluation in 2011-13.
In a further statistical analysis, the researchers compared the identified proteins with data from past studies of genetic links to Alzheimer's. The comparison suggested strongly that one of the identified proteins, SVEP1, is not just an incidental marker of Alzheimer's risk but is involved in triggering or driving the disease. SVEP1 is a protein whose normal functions remain somewhat mysterious, although in a study published earlier this year it was linked to the thickened artery condition, atherosclerosis, which underlies heart attacks and strokes. Other proteins associated with Alzheimer's risk in the new study included several key immune proteins - which is consistent with decades of findings linking Alzheimer's to abnormally intense immune activity in the brain.