MIA and CD163 as promising diagnostic biomarkers in vascular dementia: A multi-method study combining WGCNA, machine learning with validation in animal models and clinical samples

Vascular dementia (VaD), the second most common form of dementia, lacks reliable biomarkers for early diagnosis. Here, we integrated weighted gene co-expression network analysis (WGCNA) with machine learning to identify novel biomarkers and immune-metabolic pathways in VaD. Analysis of the GSE122063 dataset revealed 288 differentially expressed genes (DEGs), with four hub genes (MIA, CD163, OPALIN, SNX31) prioritized by LASSO regression and Random Forest. A nomogram model incorporating these genes achieved an AUC of 0.924, demonstrating high diagnostic accuracy. External validation (GSE186798) and experimental studies confirmed significant upregulation of MIA and CD163 in VaD patients and a 2VO rat model (P < 0.05), while SNX31 and OPALIN showed inconsistent significance. Mechanistically, MIA and CD163 correlated with macrophage polarization (M1/M2) and dysregulated oxidative phosphorylation pathways, suggesting their dual roles in neuroinflammation and metabolic reprogramming. Serum ELISA further validated elevated MIA (14.34 ± 6.32 vs. 5.23 ± 4.89 ng/mL) and CD163 (141.31 ± 71.27 vs. 58.09 ± 54.31 ng/mL) in VaD patients (P < 0.05). Our study not only establishes MIA and CD163 as robust diagnostic biomarkers but also highlights their potential as therapeutic targets for modulating immune-metabolic crosstalk in VaD pathogenesis.