Protein shape, not quantity, may be the better blood signal for Alzheimer's

The proteins were there all along. Floating in the bloodstream, in the same concentrations that standard tests already measure. But something about their shape was different in people with Alzheimer's disease, and nobody had been looking at shape.

Scientists at Scripps Research have now looked, and what they found could shift how we think about blood-based diagnostics for neurodegeneration. Their study, published in Nature Aging in February 2026, reports that structural alterations in three plasma proteins can distinguish cognitively normal individuals from those with Alzheimer's and mild cognitive impairment (MCI) with striking accuracy.

Measuring folds instead of amounts

Current blood tests for Alzheimer's typically measure how much of certain proteins, primarily amyloid beta and phosphorylated tau, are circulating. These tests have improved considerably in recent years, but they capture only part of the biological picture. The Scripps team, led by senior author John Yates and co-author Casimir Bamberger, pursued a different angle: proteostasis, the cellular machinery that keeps proteins properly folded and disposes of damaged ones.

Alzheimer's has long been associated with protein misfolding in the brain, most visibly in the amyloid plaques and tau tangles that define the disease pathologically. But proteostasis is not a brain-only system. If the protein quality-control network is failing centrally, the researchers hypothesized, signs of that failure might appear in blood proteins too.

To test this, the team analyzed plasma samples from 520 people divided into three groups: cognitively normal adults, individuals with MCI, and patients diagnosed with Alzheimer's. Using mass spectrometry, they measured how exposed or buried specific amino acid sites were on blood proteins, an indicator of whether the protein's three-dimensional structure had shifted.

Three proteins, three sites, high accuracy

Among hundreds of candidate proteins, three stood out. C1QA is involved in immune complement signaling. Clusterin plays roles in protein folding and amyloid clearance. Apolipoprotein B transports fats in the bloodstream and contributes to vascular health. At specific lysine residues on each of these proteins, structural differences tracked consistently with disease status.

A machine-learning model built on these three structural markers classified individuals across all three diagnostic categories with approximately 83 percent overall accuracy. In head-to-head comparisons between two groups, such as cognitively normal versus MCI, accuracy exceeded 93 percent.

The pattern was consistent: as Alzheimer's progressed, certain blood proteins became structurally less open. This progressive structural tightening provided a stronger diagnostic signal than simply measuring how much of each protein was present. The three-marker model performed reliably across independent cohorts and in follow-up samples collected months later, where it classified disease status with about 86 percent accuracy.

Correlation with cognition and brain structure

The structural score did not exist in a clinical vacuum. It correlated strongly with cognitive test scores and showed moderate correlation with MRI measures of brain atrophy. In repeat samples from the same individuals over time, the structural markers reflected changes in diagnostic status, suggesting the test could potentially track disease progression rather than merely capturing a snapshot.

These features matter because a useful biomarker needs to do more than separate patients from controls in a research study. It needs to move with the disease, respond to treatment effects, and remain stable enough to be clinically reproducible.

Where the evidence falls short

The study involved 520 participants, a meaningful but moderate sample size for a diagnostic validation study. Larger, multi-site trials with diverse populations would be needed before any clinical deployment. The cohorts studied were primarily from academic medical centers, and whether the structural markers perform equally well across different ethnic groups, age ranges, and clinical settings is unknown.

The mass spectrometry techniques used to measure protein structure are sophisticated laboratory methods, not bedside tests. Translating the approach into a practical clinical assay would require either significant simplification of the measurement platform or development of alternative detection methods that preserve the structural information.

The study also cannot establish causality. It is unclear whether the structural changes in blood proteins contribute to disease progression, result from it, or simply co-occur. The biological link between peripheral protein misfolding and central neurodegeneration, while plausible, remains theoretical.

Beyond Alzheimer's

The researchers are already exploring whether the same structural profiling approach could work for other diseases characterized by protein misfolding, including Parkinson's disease and certain cancers. If proteostasis decline leaves detectable structural fingerprints in the blood across multiple conditions, the diagnostic implications could extend well beyond neurology.

For now, the study offers a conceptual reframe. We have been counting proteins when we should have been watching them fold.