The Protein Shapes That Could Reveal Alzheimer’s, Years Before Symptoms Appear

Every protein in your bloodstream has a shape. Not just a composition, a shape: a tightly folded three-dimensional structure that changes depending on the molecular environment around it. In a healthy cell, chaperones and surveillance proteins keep that folding precise, correcting errors, dismantling anything that goes wrong. In the aging brain, that system starts to slip. And now researchers at Scripps Research in La Jolla believe the evidence of that slippage is detectable, long before any memory falters, in the blood.

Their new study, published in Nature Aging, proposes a different kind of Alzheimer’s blood test. Where current diagnostics measure how much of a protein is present, this one asks a subtler question: what shape is it in?

The logic starts with proteostasis, the cellular machinery responsible for keeping proteins properly folded and clearing out damaged ones. This system degrades with age, and in Alzheimer’s disease the decline is pronounced. Proteins that should hold their structure begin to sag and close in. The brain bears the brunt of it, most visibly in the amyloid plaques and tau tangles that are the disease’s hallmarks, but proteostasis isn’t a local phenomenon. If the system is failing in the brain, the researchers reasoned, might similar structural changes appear in proteins circulating in the bloodstream? “Many neurodegenerative diseases are driven by changes in protein structure,” says John Yates, a professor at Scripps Research who led the work. “The question was, are there structural changes in specific proteins that might be useful as predictive markers?”

To find out, his team turned to a technique called covalent protein profiling, or CPP. It works by tagging exposed amino acid sites on protein surfaces with chemical labels; buried sites don’t get tagged. Compare the labeling pattern in a healthy person to one in an Alzheimer’s patient and you’re essentially reading the protein’s shape rather than counting its molecules.

They ran the method on plasma samples from 520 participants across three groups: cognitively healthy adults, people with mild cognitive impairment (MCI), and people with Alzheimer’s. The results showed a consistent trend. As the disease progressed, proteins in the blood became less structurally open. The accessibility scores, averaged across hundreds of labeled sites, dropped from about 93% in healthy individuals to 92% in MCI and 91% in full Alzheimer’s. Small differences, perhaps, but statistically significant and pointing in one direction.

Out of 373 proteins profiled, three did the diagnostic work best. C1QA, a component of the immune system’s classical complement pathway. Clusterin, a protein involved in amyloid clearance that’s also the third greatest known genetic risk factor for Alzheimer’s after APOE and BIN1. And apolipoprotein B, which transports fats in the blood and has a known role in vascular health. Structural changes in specific sites on all three, tracked simultaneously and fed into a deep-learning algorithm, could classify a person as healthy, MCI, or Alzheimer’s with 83% accuracy. In head-to-head comparisons, distinguishing healthy individuals from those with MCI, or MCI from Alzheimer’s, the model’s area under the curve exceeded 0.93. By standard diagnostic benchmarks, that’s good.

“The correlation was amazing,” says Casimir Bamberger, a senior scientist at Scripps and co-author on the study. “It was very surprising to find three lysine sites on three different proteins that correlate so highly with disease state.”

What makes the finding stranger still is that the three proteins weren’t chosen because they looked similar or belonged to the same biological pathway. C1QA sits in the immune system; clusterin bridges protein folding and amyloid biology; ApoB is a lipid carrier. They share almost nothing in terms of function. The fact that structural changes in all three track with disease stage suggests the team has found something real about the global effect of proteostasis breakdown, rather than a chance association with one protein’s idiosyncratic behavior.

The panel also held up in longitudinal samples. Fifty participants were followed for up to 255 days, and the three-marker score correctly classified their disease status at follow-up in 86% of cases, including tracking some who transitioned from healthy to MCI or from MCI to Alzheimer’s during the study period. It correlated strongly with two standard cognitive tests, the MMSE and CDRSUM, and showed moderate agreement with cerebrospinal fluid amyloid and tau measurements. The paper notes that the structural panel and the CSF markers are measuring related but different things, which may turn out to be a feature rather than a limitation; complementary signals can sometimes do more than either can alone.

There are genuine caveats. The study’s longitudinal arm is modest in size, and 255 days is a short window for tracking a disease that unfolds over decades. The sample preparation method depletes the 14 most abundant proteins in plasma, which may mean some relevant signals are inadvertently removed. Larger validation cohorts and longer follow-up periods are needed before this goes anywhere near a clinic. The researchers are clear about that.

But the underlying logic is what makes this worth watching. The field has spent a generation chasing the same handful of molecular quantities, amyloid, tau, a few lipoproteins. This is asking a different question entirely. Shape, not quantity. “Detecting markers of Alzheimer’s early is absolutely critical to developing effective therapeutics,” says Yates. “If treatment can start before significant damage has been done, it may be possible to better preserve long-term memory.”

Whether the same structural-profiling approach extends to other diseases is a question the team is now beginning to explore. Parkinson’s and cancer are both on the list. The idea that the shape of a protein in your bloodstream might encode a medical future that hasn’t arrived yet is, at minimum, a compelling one to test.

DOI / Source: https://doi.org/10.1038/s43587-026-01078-2 (Nature Aging, 27 February 2026); EurekAlert press release, Scripps Research, 9 March 2026

Frequently Asked Questions

Why does protein shape matter for diagnosing Alzheimer’s, not just protein levels? Protein levels change relatively slowly across disease stages, and many proteins fluctuate for reasons unrelated to Alzheimer’s. Shape is a more direct readout of what’s going wrong inside the cell: when the machinery that keeps proteins correctly folded begins to fail, the structural changes precede other downstream effects. The Scripps team found that structural data classified disease stage with 83% accuracy, compared to only about 65% when the same proteins were assessed by abundance alone.

How is this different from the Alzheimer’s blood tests that already exist? Current blood tests for Alzheimer’s mostly measure concentrations of specific proteins, particularly amyloid beta fragments and phosphorylated tau. The Scripps approach measures whether proteins are physically opening or closing at particular sites on their surface, a distinct kind of information. The researchers suggest the two approaches may be complementary rather than competing, since they’re picking up different aspects of the same underlying biology.

Could this test be used to diagnose Alzheimer’s now? Not yet. The study involved 520 participants and followed a subset for less than a year, which isn’t large enough or long enough to establish the clinical reliability a diagnostic test requires. The research team is explicit that larger validation studies with longer follow-up periods are needed before this could be used in a clinical setting. The immediate value is in understanding what’s happening biologically.

Why do three proteins from completely different biological systems all track together in Alzheimer’s? The researchers think this reflects the systemic nature of proteostasis collapse: when the cellular machinery responsible for maintaining protein structure fails, it doesn’t fail in just one pathway. C1QA (immune system), clusterin (protein clearance and amyloid biology), and apolipoprotein B (lipid transport) share almost nothing functionally, but all three show structural changes that correlate with disease stage. That cross-system signal may be more diagnostically stable precisely because it’s unlikely to be confounded by disruption to any single pathway.

What would need to happen for this to become a routine blood test? The next steps would involve validating the three-protein panel in much larger, more diverse cohorts, ideally including people who are followed from before any symptoms appear through to diagnosis. The study would also need to demonstrate that the test performs consistently across different labs using different equipment. If those hurdles are cleared, the relative simplicity of a blood draw compared to a PET scan or lumbar puncture makes the clinical case compelling.