Sleep-wake disturbances commonly occur in Alzheimer's disease (AD). However, the precise mechanisms underlying the breakdown of the circadian gene network remain elusive. This study examined 23 circadian genes in 253 post-mortem samples (80 AD, 173 controls) from the hippocampus, entorhinal cortex, and superior frontal cortex, using three metrics: a Circadian Disruption Index (CDI), a Network Coherence Score (NCS), and a Circadian Entropy Index (CEI). AD brains showed a modest overall shift in circadian gene expression (CDI = 0.101), concentrated mainly in CRY2, PER3, and PER2. The PER3-CRY2 gene pair showed the most significant co-expression alteration of any gene pair examined, a 60.4% reduction in coherence (∆ρ 0.604). Network-wide coherence (NCS) and CEI, averaged across all 23 genes, did not differ significantly between the AD and control groups (NCS: 6.7% reduction, p = 0.356; CEI: -0.4% change, p = 0.923), suggesting the alteration is concentrated in specific gene relationships rather than global network collapse. Regional CDI did not differ significantly across brain areas (p = 0.724). A 10-gene panel classified AD versus control with cross‑validated area under the curve (AUC) = 0.759 ± 0.069 (nested cross-validation AUC = 0.759 ± 0.069, zero optimism bias), though external validation in an independent cohort (GSE5281) did not exceed chance level (AUC = 0.500). These results identify the PER3-CRY2 co-expression change as a specific, reproducible molecular signature of AD and support a validated circadian gene-based classifier, while showing that global circadian network disruption is not a consistent feature of this cohort. The findings identify a localized alteration in circadian gene co-expression in AD brain tissue, with potential implications for future therapeutic targeting of the PER3-CRY2 interaction.
