Blood protein structure changes may enable earlier detection of Alzheimer’s

By Dr. Priyom Bose, Ph.D.Reviewed by Lauren HardakerMar 13 2026

Scientists combine advanced proteomics and AI to reveal structural blood protein changes that could help distinguish early Alzheimer’s from mild cognitive impairment.

Study: Structural signature of plasma proteins classifies the status of Alzheimer’s disease. Image credit: Beyond This/Shutterstock.com

A recent study in Nature Aging combined mass spectrometry-based structural proteomics with machine learning to establish a minimally invasive, reliable, and potentially scalable research strategy for early detection and classification of Alzheimer’s disease (AD) and related cognitive conditions.

Protein homeostasis disruption and structural biomarkers in Alzheimer’s disease

Proteostasis, or protein homeostasis, refers to the cellular processes that maintain proper protein folding, stability, and degradation. These mechanisms are crucial because a substantial proportion of newly synthesized proteins can misfold, disrupting normal cell function if not managed by cellular quality control systems.

In AD, the machinery responsible for proteostasis becomes less effective, allowing misfolded proteins and damaged cellular components to build up over time. This impaired clearance supports the early accumulation of amyloid-β aggregates, abnormal protein clumps that can form in the brain years before the first signs of Alzheimer’s symptoms appear. A comprehensive understanding of protein conformational changes and interactions, beyond the traditional focus on amyloid plaques and tau tangles, could uncover disease mechanisms and plasma-based structural biomarkers.

Apolipoprotein E (APOE) is a polymorphic plasma protein with three major isoforms (ε2, ε3, ε4) differing by one or two amino acids, leading to altered binding properties. The ε4 allele is strongly associated with increased AD risk, while ε2 confers protection. Despite extensive characterization of APOE genotype expression profiles and network effects, the impact of APOE variants on the structure of ApoE-interacting proteins remains underexplored.

Neuropsychiatric symptoms (NPSs) are prevalent in AD, with sex differences noted in progression and symptomatology. Women tend to experience more rapid cognitive decline and higher rates of delusion, while men exhibit increased apathy and agitation. Despite growing efforts to define molecular correlates of NPSs, the relationship between sex and NPSs remains unclear due to clinical heterogeneity in AD.

Assessing protein structure alterations in Alzheimer’s disease

Blood samples were collected from participants at the University of California, San Diego (UCSD) and the University of Southern California Alzheimer’s Disease Research Centers. Alzheimer’s pathology in the UCSD cohort was supported by cerebrospinal fluid (CSF) measurements of amyloid-β and tau, while clinical status across cohorts was evaluated using established diagnostic criteria. Participants were assessed biannually for cognitive function and categorized using standard criteria, including the Clinical Dementia Rating (CDR) and neuropsychological testing.

Peptide samples were analyzed by Liquid Chromatography–Tandem Mass Spectrometry (LC–MS/MS) coupled to a timsTOF Pro mass spectrometer. A machine learning framework was used to classify mass spectrometry data, with a deep neural network selected after benchmarking against 17 additional machine learning algorithms.

Identification of structural blood biomarkers for early AD detection

A total of 520 blood samples were obtained from two large cohorts. By combining blood assessment findings with detailed clinical and biomarker data, including cognitive tests and cerebrospinal fluid (CSF) measures, where available, researchers classified Alzheimer’s disease (AD) status and progression.

Notably, both cohorts were well-matched for age, and APOE genotyping was performed. As expected, cognitive scores such as the MMSE (Mini-Mental State Examination) and CDR-SUM (Clinical Dementia Rating Sum of Boxes) reliably tracked with increasing disease severity, anchoring the study in established clinical metrics.

Covalent protein profiling (CPP) was performed to measure lysine residue accessibility in blood proteins, which serves as a proxy for protein conformational state. This technique detected subtle structural changes in proteins, a novel approach, since most studies focus only on protein quantity. The authors focused on abundant proteins for potential clinical translation.

Interestingly, the current study observed that alterations in protein structure, rather than abundance, correlated more strongly with AD. As the disease progressed, proteins showed less accessible lysines and greater variability, indicating that loss of protein homeostasis may represent an important molecular feature of Alzheimer’s. Protein structural changes were not always linear, with some appearing early in the disease and others later. This nonlinearity suggests that protein conformation shifts could serve as sensitive indicators, potentially detecting disease before traditional markers.

APOE ε4 was associated with lower protein accessibility in several proteins, most notably C1QA and SERPINA3. Computational modeling suggested that these changes could reflect altered protein–protein interactions, offering insights into how genetics might shape protein structure in AD. Computational modeling provided structural support for these experimentally observed associations, suggesting that the APOE genotype may influence structural changes in circulating proteins and thereby shedding light on the molecular underpinnings of AD risk.

The current study also found that decreased protein accessibility tracked with NPS severity, and several proteins showed sex-specific associations, many linked to amyloidosis and established AD pathways. NPS scores were more diagnostically powerful in women, and proteins, such as CLUS and ITIH2, exhibited notable sex-specific structural changes that mirrored disease stage and symptom burden. This finding suggests that sex-informed biomarker patterns could potentially improve diagnostic precision, although further investigation is needed.

A diagnostic panel for AD diagnosis

The authors developed a multi-marker panel comprising C1QA, CLUS, and ApoB, using machine learning. Their deep learning model achieved approximately 83 % accuracy in distinguishing healthy, MCI, and AD cases, significantly surpassing models based solely on protein abundance.

Most misclassifications were limited to adjacent disease stages, showing the panel’s nuanced discrimination. The robustness of the panel was further demonstrated, with accuracy holding steady even with data imputation across different cohorts. The authors also reported that age did not significantly confound the performance of the multi-marker panel, supporting the reliability and potential generalizability of the findings.

The structural markers demonstrated strong correlations with cognitive scores and moderate associations with brain imaging measures, including MRI-derived indices associated with Alzheimer’s pathology, and CSF biomarkers, confirming their clinical and pathological relevance. Notably, these blood-based measures offer a less invasive alternative to existing diagnostics.

In longitudinal analysis, the marker panel tracked disease progression with about 86 % accuracy within the study’s relatively short follow-up window and showed changes that corresponded with shifts in diagnostic status, highlighting its promise for monitoring ongoing physiological changes in AD.

Conclusions

The current study demonstrates that a blood panel assessing structural changes in the C1QA, CLUS, and ApoB proteins provides a promising experimental biomarker approach for diagnosing and tracking AD. By prioritizing protein conformation rather than abundance, the approach paves the way for less invasive diagnostics that could support earlier detection strategies if validated in larger, prospective, and longer-term studies.

Download your PDF copy by clicking here.