Alzheimer’s disease (AD) is a destructive neurodegenerative disorder with a rising global incidence, imposing a significant economic burden on healthcare systems (Scheltens et al., 2021; Zheng and Wang, 2025). As the predominant form of dementia, AD manifests as progressive cognitive decline, memory impairment, and behavioral alterations (Mangal and Ding, 2022; Liu et al., 2024). Mounting evidence suggests that neuropathological changes precede clinical symptoms in AD patients by several decades (Chen and Zhang, 2022; Xu et al., 2024). Therefore, a comprehensive understanding of AD pathogenesis and the identification of novel genes for early diagnosis and treatment are urgent issues that demand attention.

The circadian rhythm serves as an intrinsic biological clock in the human body, forming an approximately 24-h cycle (Xiao et al., 2023). Disruption of the circadian rhythm is a common characteristic observed in neurodegenerative disorders such as AD (Amidfar et al., 2023; Leng et al., 2019; Chen et al., 2024). Impairments in circadian rhythm and sleep patterns among AD patients include fragmented sleep, heightened arousal at night, and reduced daytime activity levels, which can exacerbate AD-related pathologies (Zhang et al., 2025). Studies in mice and humans have shown fluctuations in levels of Aβ and tau proteins in interstitial or cerebrospinal fluid during the sleep/wake cycle, with peaks during the active phase (Kang et al., 2009; Holth et al., 2019; Ma et al., 2016; Roh et al., 2014; Xie et al., 2013; Nguyen Ho et al., 2024). Circadian disruption has also been associated with immune system dysregulation, which can lead to neuronal injury and cognitive decline in AD (He et al., 2023; Weng et al., 2024). Because the circadian rhythm is regulated by a transcription–translation feedback loop involving clock genes (Schrader et al., 2024), identification of clock gene alterations is crucial for further investigations into AD pathogenesis. Therefore, we aim to explore the relationship between clock genes and AD through bioinformatics analysis in the study.

Recently, research on AD has focused on multimodal neuroimaging and deep learning frameworks (Upadhyay et al., 2024b). Upadhyay et al. conducted a comprehensive review of deep learning methods for AD classification, highlighting the potential for integrating structural MRI (sMRI) and functional MRI (fMRI) data (Upadhyay et al., 2024a). Similarly, a study proposed a 3D deep learning model for early AD diagnosis using sMRI and PET imaging, achieving an accuracy of 91.84% (Raza et al., 2024). These new methods can improve the accuracy of AD diagnosis. In contrast to these approaches, our study uniquely combines machine learning with circadian biology to identify biomarkers associated with clock genes. Instead of previous studies that mainly relied on neuroimaging data, we used gene expression profiles to discover new diagnostic markers. This comprehensive approach not only improves diagnostic accuracy but also provides insights into the molecular mechanisms underlying AD progression.

In this study, we utilized a variety of bioinformatics methods and machine learning algorithms to explore the clock genes in AD, with the goal of identifying potential early predictive biomarkers for AD patients and understanding the pathological mechanisms involved. The findings of this study can potentially facilitate the discovery of diagnostic markers for AD. In addition, compared to previous studies (Bacalini et al., 2022; Niu et al., 2022) focusing on single-clock genes (e.g., BMAL1 or PER1), our study integrated multiple machine learning algorithms to prioritize biomarkers (ERC2, PRKCG, and HLA-DMA) with higher diagnostic specificity (AUC > 0.7).

Alzheimer’s disease is one of the most commonly diagnosed cases of senile dementia in the world (Twarowski and Herbet, 2023). Data from the Alzheimer’s Association Report revealed that approximately 50 million individuals worldwide were afflicted by AD in 2018, with projections estimating a rise to 152 million by 2050 (Zhang et al., 2022). The elevated prevalence of AD imposes a substantial economic and societal burden, presenting a formidable challenge for both individuals and communities (Huang et al., 2024). Current therapeutic approaches to AD encompass cognitive enhancement therapy, management of neuropsychiatric symptoms, and disease-modifying treatments (Zhang et al., 2023). However, several drugs are still under investigation and exhibit limited efficacy (Thakkar et al., 2023). Therefore, exploring new treatment strategies is an urgent task for researchers.

Circadian rhythms, intrinsic 24-h oscillations in charge of regulating daily behaviors and biological processes (Jin et al., 2023), have garnered attention concerning their potential impact on AD. Although the causal relationship between Alzheimer’s disease and circadian rhythm disruption is unclear, there is growing evidence that AD and circadian disruption may interact, with circadian disruption increasing the risk of AD development and AD exacerbating circadian disruption (Rigat et al., 2023; Uddin et al., 2020; McKee et al., 2020). Notably, clock genes have become pivotal roles in regulating circadian rhythms, attracting increased research interest (Zhu et al., 2022). Circadian rhythms are influenced by the dark–light cycle and are chiefly governed via a series of molecular clocks in the suprachiasmatic nucleus (Uddin et al., 2021). Hub clock genes, such as period circadian protein homologs, cryptochromes (CRY1 and CRY2), CLOCK, and BMAL1, coordinate an intricate transcription–translation feedback loop oscillating each 24 h (Takahashi, 2017; Vitaterna et al., 2001). Many studies have highlighted the intimate relation between clock genes and AD pathogenesis (Lananna et al., 2020; Buhl et al., 2019).

In the study, we recognized 560 DEGs between AD and ND groups, with 323 upregulated and 237 downregulated. Through Venn analysis, nine intersection genes (PRKCG, TUBB3, ADCY1, GABBR2, ERC2, GSTM3, TBL1X, HLA-DMA, and CLDN5) were identified from the DEGs and clock genes. These findings hold promise as a foundation for subsequent animal experiments or investigations using human samples, offering valuable insights into the interplay between clock genes and AD progression.

EA of DECGs showed high enrichment levels in BP, including synaptic vesicle, exocytosis neurotransmitter, signal release from synapse, synaptic vesicle cycle, regulation of synaptic plasticity, vesicle-mediated transport in synapse, regulation of neurotransmitter levels, regulated exocytosis, and learning or memory. These processes are closely associated with neurotransmitters, synapses, and learning or memory. The enriched cell components displayed that DECGs were prevalent in the synaptic membrane and distal axon. In terms of molecular function, differentially expressed clock genes are mainly enriched in glutathione binding, structural constituents of synapse, and GABA receptor activity.

KEGG enrichment analysis was mainly enriched in gap junction, GABAergic synapse, long-term potentiation, thyroid hormone synthesis, insulin secretion, melanogenesis, pancreatic secretion, parathyroid hormone synthesis, secretion and action, cholinergic synapse, leukocyte transendothelial migration, glutamatergic synapse, and so on. Notably, in addition to circadian entrainment and synapse-related pathways, other pathways are associated with metabolism and neurotransmission. The high energy requirements of the brain make it very sensitive to modifications in energy metabolism, and metabolic disorders are a hallmark of brain aging, especially in neurodegenerative disorders, for instance, AD. AD patients are often associated with metabolic abnormalities, such as abnormal insulin secretion and parathyroid hormone synthesis (Chen et al., 2023; Patel and Edison, 2024). These metabolic abnormalities further aggravate AD by affecting brain energy metabolism and neuronal health. For example, in AD patients, excessive release of glutamate and overactivation of NMDA receptors lead to increased calcium influx, which in turn causes metabolic disorders (Le Douce et al., 2020). The results of this study emphasize the correlativity between clock genes associated with AD and metabolism and further reveal the correlativity between AD and metabolic disease from the view of clock genes.

Four machine learning methods (LASSO, random forest algorithm, SVM-RFE, and GMM) were utilized to identify differential expression clock genes, and four key genes (GSTM3, ERC2, PRKCG, and HLA.DMA) were obtained. In addition, the expression levels of four genes were evaluated in training and validation datasets, and a clock gene diagnosis model for AD was established and validated. In the construction of the AD model, the three most relevant clock genes were ERC2, PRKCG, and HLA-DMA. Notably, we built a more comprehensive diagnostic nomogram model on account of three hub genes in training and validation datasets, which is more valuable for the diagnosis of AD than independent biomarkers.

We further analyzed the specific signaling pathways of the three hub genes and explored the underlying molecular mechanisms by which they affect AD progression. Single-gene GSEA results show amyotrophic lateral sclerosis, oxidative phosphorylation, Parkinson’s disease, pathways of neurodegeneration-multiple diseases, cardiac muscle contraction, endocrine and other factor-regulated calcium reabsorption, and synaptic vesicle cycle. Previous studies have shown that neurological diseases are closely associated with circadian rhythms. In our study, clock genes associated with AD were concentrated in the pathways related to neurological and metabolic dysregulation, further proving a shared pathway between AD and these systemic disorders from a circadian view.

Previous studies have shown that some clock genes can cause AD through the immune system (Li et al., 2020; Song, 2019). In this study, we found significant differences among activated B cells, activated CD8 T cells, immature B cells, memory B cells, mast cells, MDSC, natural killer cells, natural killer T cells, neutrophils, and other immune cells between AD patients and healthy individuals. HLA-DMA showed a positive correlation with neutrophils, mast cells, T follicular helper cells, effector memory CD8 T cells, natural killer cells, MDSC, natural killer T cells, and CD56dim natural killer cells. However, ERC2 and PRKCG were negatively correlated with neutrophils, mast cells, T follicular helper cells, effector memory CD8 T cells, natural killer cells, MDSC, natural killer T cells, and CD56dim natural killer cells. In summary, these three hub genes exhibited a complex relationship with immune cell infiltration in AD, indicating their potential role in maintaining the balance of the immune response. Therefore, a deeper understanding of the immune mechanisms may enhance the diagnosis of AD and facilitate the development of effective treatments.

This study forecasts potential therapeutic agents for clock genes closely associated with AD patients. These agents include circadian rhythm modulators: bryostatin and bryostatin 1; anti-inflammatory and immunomodulator agents: fasudil, SB220025, UCN-01, PF-562271, tamatinib, and SP-600125; neuroprotective agents: lauroguadine, ingenol mebutate, and quercetin. These potential agents targeting clock genes treat AD through different mechanisms, providing a theoretical basis and feasible direction for subsequent relevant research.

The miRNA plays a key part in the modulation of gene expression and influences the post-transcriptional regulation of genes by binding to mRNA (Fabian et al., 2010). Research has indicated that miRNAs play a crucial part in neurological diseases, including AD (Klyucherev et al., 2022). In our study, we explored the clock genes associated with AD and their corresponding miRNA regulatory networks, revealing the potential mechanisms of these genes in AD pathology. By establishing a miRNA–mRNA network, we obtained miRNAs associated with three key clock genes (HLA-DMA, PRKCG, and ERC2). The interaction of these miRNAs with clock genes may play a crucial part in the occurrence and progress of AD. The construction of miRNA regulatory networks provides a novel perspective for comprehending the intricate gene regulatory mechanism of AD.

EA of 576 miRNAs displayed that they were enriched in β1 integrin cell surface interactions, TRAIL signaling pathway, etc. These pathways play a crucial part in neurodegenerative and metabolic diseases, suggesting that these miRNAs may take part in the pathological course of AD by influencing these key pathways. This study offers important clues for further exploring the specific mechanism of miRNA in AD and is helpful for the future development of miRNA treatment methods for AD. Although this study reveals a potential mechanism of action for miRNAs in AD, further experiments are needed to validate these findings. Future studies can verify the specific regulatory relationship between these miRNAs and clock genes in AD through in vivo and in vitro experiments. In addition, the in-depth exploration of miRNA expression patterns at different stages of AD and in different patient populations will help to develop personalized treatment regimens.

This study had the following three advantages: (1) This is a novel integration of four machine learning algorithms (LASSO, RF, SVM-RFE, and GMM) to identify AD biomarkers from clock genes, addressing limitations of single-model approaches; (2) It is the first time to combine HLA-DMA, PRKCG, and ERC2 to form a nomogram for AD diagnosis, thus achieving higher diagnostic value; and (3) We systematically explored the role of circadian immune crosstalk in AD. However, this study also had some limitations. First, the dependency on public datasets can induce biases associated with sample collection and management. Second, the variability in gene expression trends across datasets underscores the need for larger, more diverse cohorts to enhance the robustness of our results.

Future studies should emphasize the experimental validation of the identified hub genes and their associated pathways in AD models. Investigating the molecular mechanisms through which these genes influence circadian rhythms and metabolism could provide deeper insights into AD pathology. In addition, exploring the therapeutic potential of the identified drugs in preclinical and clinical studies is crucial. Expanding the scope of immune infiltration analysis to include a broader range of immune cells and their interactions with clock genes may also uncover new therapeutic targets.

In conclusion, we identified three clock gene-related biomarkers (ERC2, PRKCG, and HLA-DMA) with high diagnostic performance for AD through four machine learning algorithms (LASSO regression, Random Forest, SVM-RFE, and GMM). Functional enrichment analysis revealed that these biomarkers were involved in key pathways such as oxidative phosphorylation and neurodegeneration, highlighting their potential roles in AD pathogenesis. Immune infiltration analysis further uncovered significant differences in 15 immune cell subsets between AD and control groups, suggesting a link between circadian disruption and neuroinflammation. In addition, we identified 23 drugs targeting HLA-DMA and 8 drugs targeting PRKCG, providing a foundation for future therapeutic development. The mRNA–miRNA regulatory network analysis offered novel insights into the post-transcriptional mechanisms underlying AD. These findings not only advance our understanding of the molecular mechanisms linking circadian rhythms to AD but also provide valuable tools for early diagnosis and targeted therapy. Future studies should focus on the experimental validation of these biomarkers and their therapeutic potential in preclinical models.