Brain-derived 5-hydroxymethylcytosine epigenetic scores are related to Alzheimer’s disease pathology and cognitive decline

Oct 28, 2025Acta neuropathologica

Brain levels of 5-hydroxymethylcytosine linked to Alzheimer's disease and memory loss

AI simplified

Alzheimers & Treatments on OpenScience ↗PubMed ↗DOI ↗OA ↗

Abstract

AUC values of 87.0% in the training set and 91.4% in the validation set indicate high performance in distinguishing Alzheimer's disease from non-Alzheimer's pathology using brain-derived epigenetic scores.

Brain-derived (5hmC) profiles were analyzed from 1005 postmortem human brain samples.
655 participants were classified as having Alzheimer's disease (AD) based on neuropathologic criteria.
Machine learning models effectively distinguished AD from non-AD pathology using 136 candidate gene bodies and 96 enhancers.
Pathway analyses suggested involvement of cardiovascular function, endocytosis, and MAPK signaling pathways in AD.
The developed AD-score was significantly associated with rates of cognitive decline across multiple domains.

AI simplified

BACKGROUND: Cytosine modifications play critical roles in gene regulation and disease pathogenesis. Elucidating novel epigenetic contributions to Alzheimer's disease (AD) could advance diagnostic, prognostic, and therapeutic strategies. (5hmC), a stable and dynamic DNA modification, has emerging links to AD pathophysiology and potential uses as a biomarker. Cognitive decline, a hallmark of AD progression, varies across individuals and is not fully explained by classic pathological markers. Here, we aimed to develop and validate brain-derived 5hmC-based epigenetic scores to distinguish AD from non-AD pathology and examine their relationship to individual cognitive trajectories.

METHODS: Genome-wide 5hmC profiles were generated using 5hmC-Seal and next-generation sequencing on 1016 postmortem human brain prefrontal cortex samples from well-characterized, deceased participants in a longitudinal, clinical-pathologic research study on aging. Samples were processed in independent training and validation sets. Genomic features (e.g., gene bodies, enhancers) were summarized, followed by differential analysis and machine learning-based feature selection to construct classification models distinguishing AD from non-AD.

RESULTS: After quality control and batch correction of 5hmC data, 1005 participants were included, with 655 classified as having AD and 350 as non-AD, according to NIA-AA neuropathologic criteria. In the training set (n = 859), 136 candidate gene bodies and 96 enhancers were selected based on variability and relaxed significance thresholds (p < 0.1). Pathway enrichment analyses implicated cardiovascular function, endocytosis, and MAPK signaling pathways. Using these features, we developed machine learning models that distinguished AD from non-AD with high performance in both the training set (AUC = 87.0%; 95% CI 84.2-89.7%) and validation set (n = 146; AUC = 91.4%; 95% CI 86.6-96.2%). Moreover, the resulting AD-score was significantly associated with rates of global and five domain-specific cognitive decline.

CONCLUSION: This study extends prior work by translating brain 5hmC profiles into epigenetic scores that distinguish AD pathology and reflect individual cognitive trajectories. These findings highlight the potential of brain-derived 5hmC modifications as biomarkers for AD and as tools to advance research on disease progression, offering a new direction for epigenetics-informed clinical applications in AD.

Key numbers

91.4%

(Validation Set)

Area under the curve for the validation set classification model.

78.3%

Sensitivity of Model

Sensitivity achieved in the validation set for the -based model.

Full Text

We can’t show the full text here under this license. Use the link below to read it at the source.

View full text (PDF) ↗View full text (HTML) ↗

Brain-derived 5-hydroxymethylcytosine epigenetic scores are related to Alzheimer’s disease pathology and cognitive decline

Abstract

Key numbers

Full Text

You found one interesting study. We’ll send the next 7.

what lands in your inbox each week:

Recent issues from the alzheimers & treatments brief