Severe sleep disturbances affect the vast majority of people living with Alzheimer’s disease. Sundowning, nighttime wandering, and fragmented sleep are among the most distressing symptoms for both patients and caregivers, and they have long pointed scientists toward a simple hypothesis: the circadian clock must be broken in the Alzheimer’s brain. A new study published in Chronobiology International puts that hypothesis under a more precise microscope than ever before, and the picture that emerges is not the wholesale collapse researchers might have expected. It is, instead, a far more specific molecular failure.
Researchers from P.S.R. Engineering College and Jairam Arts and Science College in Tamil Nadu, India, examined the expression of 23 core circadian genes in 253 post-mortem brain tissue samples drawn from the hippocampus, entorhinal cortex, and superior frontal cortex of 80 Alzheimer’s donors and 173 cognitively normal controls. Rather than simply asking whether individual genes were turned up or down, the team applied three mathematical metrics designed to capture different aspects of circadian network integrity: the Circadian Disruption Index, the Network Coherence Score, and the Circadian Entropy Index. Each measures a distinct kind of disruption, and together they allowed the researchers to ask not just whether the clock is broken in Alzheimer’s, but how it is broken and at what scale.
What they found.
The headline finding is not one of broad collapse but of a single, highly specific disconnection. The overall Circadian Disruption Index across all 23 genes was modest, just 0.101 on a scale designed to detect shifts in rhythmic expression. That number suggests that, counter to what many clinicians might assume, the circadian machinery in the Alzheimer’s brain is not, on average, massively deranged.
The real signal emerged when the researchers looked at pairwise co-expression relationships between individual genes. Under normal conditions, circadian genes work in coordinated feedback loops. PER proteins dimerize with CRY proteins to form the negative arm of the molecular clock, and the strength of that partnership can be measured by how closely the two genes rise and fall together across samples. The PER3-CRY2 pairing stood out dramatically. In Alzheimer’s tissue, the co-expression between these two genes dropped by 60.4 percent relative to controls, the largest reduction of any gene pair in the study.
This is a molecular signature, not a systemic breakdown. The Network Coherence Score, which measures how well the entire 23-gene network holds together as an interconnected system, showed only a 6.7 percent reduction, a shift that did not reach statistical significance. The Circadian Entropy Index, a measure of how disordered or randomized the expression patterns have become, changed by less than 1 percent. By these global metrics, the circadian network in the Alzheimer’s brain looks surprisingly intact.
The specificity of the finding is reinforced by what happened next. The researchers built a 10-gene classification panel designed to distinguish Alzheimer’s from control samples based on the expression levels of carefully selected circadian genes. On the training data, the panel performed reasonably well, achieving an Area Under the Curve of 0.759, meaning it could correctly discriminate between the two groups about three-quarters of the time. But when the same panel was tested on an independent external dataset, its performance collapsed to an AUC of 0.500, exactly the result expected from random guessing. A model built on broad circadian gene expression profiles could not generalize. The PER3-CRY2 co-expression disruption, by contrast, was robust and replicable.
Why it matters.
The finding carries weight for several reasons. First, it reframes a clinical observation that has been treated as obvious and monolithic. Alzheimer’s patients have disrupted sleep, and brain banks show that amyloid plaques and tau tangles accumulate in regions that regulate the circadian clock. The natural inference has been that disease pathology globally damages the clock machinery. This study suggests instead that the clock architecture is largely preserved, with a single critical coupling coming undone.
That specificity matters because it changes what a pharmacological or therapeutic intervention might target. A drug designed to broadly stabilize circadian gene expression would be working against a vague and potentially nonexistent target. A therapy aimed at strengthening the PER3-CRY2 interaction, or compensating for its loss, has a defined molecular target and a measurable outcome. The 60.4 percent drop in co-expression is not subtle; it is a large, unambiguous signal embedded in a background of relative stability.
The choice of brain regions is also significant. The hippocampus and entorhinal cortex are among the earliest and most heavily affected sites of Alzheimer’s pathology, while the superior frontal cortex is more resistant to early degeneration. By sampling across all three regions, the study offers some indication that the PER3-CRY2 disruption is not simply a secondary consequence of tissue degeneration. If it were, one would expect the effect to be largest in the most damaged regions and absent in the least damaged ones. The regional pattern the authors describe suggests a more primary molecular involvement.
The results also offer a cautionary methodological note. The failure of the 10-gene panel to generalize to an external dataset is a reminder that the history of molecular biomarker discovery in Alzheimer’s is littered with promising panels that could not survive cross-validation. The circadian field would do well to learn that lesson early. A single, replicable, mechanistically interpretable signal like the PER3-CRY2 co-expression deficit may ultimately prove more valuable than a complex but unstable multivariate classifier.
Limits.
The study has important constraints. Post-mortem tissue captures a single snapshot at the end of life, not the dynamic rhythm of gene expression across a 24-hour cycle. The Alzheimer’s donors were, on average, at advanced stages of the disease, and it is possible that earlier-stage tissue would show a different pattern of circadian disruption, perhaps a more global one that later narrows to the PER3-CRY2 signature. The sample size of 80 Alzheimer’s cases, while reasonable for a post-mortem gene expression study, is not large enough to stratify by disease stage, medication history, or sleep symptom severity, all of which could influence the results. The study also did not include antemortem sleep measurements, so there is no direct way to connect the molecular finding to the behavioral sleep symptoms that make circadian disruption clinically meaningful in the first place.
Bottom line.
Alzheimer’s disease is associated with a large and specific loss of co-expression between two core circadian genes, PER3 and CRY2, rather than a general breakdown of the circadian gene network. The finding represents one of the most precisely localized molecular signatures of circadian disruption ever described in the Alzheimer’s brain. Whether it is a cause, a consequence, or a contributor to the sleep symptoms that plague patients will require prospective studies that track gene expression alongside behavior and disease progression. For now, the PER3-CRY2 pair offers a target worth watching.
Source. Venkatesh S, Venkadeswaran K, Ravi Raja, et al. Mathematical mapping of circadian genes in Alzheimer’s disease. Chronobiology International. 2026. DOI: 10.1080/07420528.2026.2705386. PMID: 42483877.