Identification of CLOCK and MARK2 as Candidate Blood-Based Biomarkers in Type 2 Diabetes: An Integrated Bulk and Single-Cell Transcriptomic Study

Abstract

Background: Type 2 diabetes mellitus (T2DM) is often accompanied by nervous system complications. However, the molecular mechanisms underlying T2DM remain incompletely understood. This study aimed to identify candidate blood-based biomarkers associated with T2DM by screening a neural-related gene set through integrated bulk and single-cell transcriptomic analysis of peripheral blood. The present analysis focused on the systemic expression patterns of the selected genes in peripheral blood as indicators of T2DM-associated immunometabolic dysregulation, rather than on neurological involvement. The neural-related gene set was used only as a predefined screening starting point. Methods: We employed a three-tier validation framework comprising (1) a training cohort (GSE15932, 8 T2DM patients vs. 8 controls) for biomarker discovery using differential expression analysis and machine learning (XGBoost, random forest, and PPI network topology); (2) an independent validation cohort (GSE184050, 50 T2DM patients vs. 65 controls) for cross-dataset diagnostic performance assessment (ROC analysis); and (3) an experimental validation cohort (5 T2DM patients vs. 5 controls) for independent clinical confirmation by RT-qPCR. Additionally, single-cell RNA sequencing (GSE268210) was used to investigate cell type-specific expression and immune regulatory networks. Results: By combining multiple screening methods and cross-dataset validation, we identified CLOCK and MARK2 as key candidate genes from the neural-related gene set. CLOCK and MARK2 showed high discriminatory performance in the discovery cohort, whereas their AUCs were 0.613 and 0.653, respectively, in the validation cohort, indicating limited standalone discriminatory ability. Both genes are highly expressed in peripheral blood monocytes and are associated with MIF-(CD74+CXCR4)-mediated cell communication, TNFA-NFκB inflammatory signaling, and activation of oxidative phosphorylation pathways. Finally, RT-qPCR in clinical samples confirmed that both genes were indeed upregulated (p < 0.05). Conclusions: CLOCK and MARK2 were identified as candidate T2DM-associated blood-based biomarkers with reproducible expression differences across datasets. Their discriminatory performance was more modest in the independent validation cohort, indicating limited standalone diagnostic utility. Single-cell and computational analyses further suggested potential immunometabolic relevance. Because neurological phenotyping, cognitive assessment, and direct measures of nervous system involvement were not available in this study, any relationship between these genes and the neurological manifestations of T2DM remains speculative and hypothesis-generating, and requires prospective validation in cohorts with matched neurocognitive data.