The Brief

Scientists at Scripps Research have demonstrated that structural changes in three blood plasma proteins — measured by how they fold rather than how much is present — can distinguish healthy individuals from those with Alzheimer’s and mild cognitive impairment with over 93% accuracy in direct comparisons. The findings, published in Nature Aging, establish an entirely new class of blood-based biomarkers targeting the breakdown of the body’s protein-maintenance system rather than the accumulation of specific pathological proteins.

The Report

A team led by senior author John Yates at the Scripps Research Institute has shown that measuring how blood proteins are shaped — rather than counting how many are present — can accurately classify stages of Alzheimer’s disease. The study, published February 27 in Nature Aging, analyzed blood plasma from 520 participants across NIA-funded Alzheimer’s Disease Research Centers in Kansas and California.

Using mass spectrometry combined with machine learning, the researchers identified structural changes at specific sites on three proteins: C1QA, involved in immune signaling; clusterin, which assists with protein folding and amyloid clearance; and apolipoprotein B, which transports fats and supports blood vessel health. As Alzheimer’s progressed, these proteins became less structurally open — a shift invisible to conventional blood tests that measure only concentration.

The three-protein panel achieved 83.44% accuracy in classifying participants across all three groups — cognitively normal, mild cognitive impairment, and diagnosed Alzheimer’s. In direct two-way comparisons, accuracy exceeded 93%: 93.43% for distinguishing healthy individuals from those with mild cognitive impairment, and 93.25% for separating mild cognitive impairment from Alzheimer’s. When the team retested participants months later, the panel held at roughly 86% accuracy and reflected changes in diagnostic status over time. The structural score also correlated strongly with cognitive test scores and moderately with MRI measures of brain atrophy.

The approach is grounded in proteostasis — the body’s system for maintaining properly folded proteins and clearing damaged ones. Alzheimer’s has long been associated with amyloid plaques and tau tangles, but growing evidence implicates a broader collapse in this maintenance system. The Scripps team’s insight was that if proteostasis is failing in the brain, the signature might be legible in proteins circulating in the blood.

“This work introduces a fundamentally new, blood-based approach to detecting and staging Alzheimer’s disease,” said Dr. Richard Hodes, director of the National Institute on Aging. The researchers position the test as complementary to existing FDA-approved blood diagnostics, which measure levels of amyloid beta and phosphorylated tau and already achieve approximately 90% accuracy. Co-author Casimir Bamberger noted surprise at the strength of the findings: “It was very surprising to find three lysine sites on three different proteins that correlate so highly with disease state.”

The study’s longitudinal cohort was small — fewer than 50 participants tracked for under a year — and the authors acknowledge that larger validation studies are required before clinical use. The team is now exploring whether the same structural approach can detect Parkinson’s disease and certain cancers. Alzheimer’s currently affects approximately 7.2 million Americans aged 65 and older.

The Angle

The distinction this study draws — between what is present and what has gone wrong with what is present — is worth more attention than the accuracy numbers alone suggest. Every existing Alzheimer’s blood test asks the same question: how much of this protein do you have? The Scripps team asked a different one: what shape are your proteins in? That shift from quantity to conformation is not an incremental improvement on an existing method. It is a different category of measurement, targeting a different layer of biological process.

What makes this consequential is the implication for timing. Current biomarkers detect the downstream products of disease — the plaques and tangles that accumulate after damage is already underway. Proteostasis breakdown is upstream. Proteins begin misfolding before pathological aggregates form. If structural signatures in the blood can be read before the conventional markers appear, the diagnostic window opens earlier — and earlier is the only direction that matters for a disease whose treatments work best when there is still something left to preserve.

The researchers are already exploring applications to Parkinson’s and cancer, both of which involve their own versions of proteostasis failure. That expansion hints at something larger than a single diagnostic: the possibility that protein shape, read cheaply from a blood draw, becomes a general-purpose early warning system for the class of diseases that kill by corrupting the body’s own maintenance machinery. The new disease-modifying drugs exist. The question has always been whether diagnosis can move fast enough to make them useful. This is a credible attempt to answer it.