Identification and Validation of the Potential Key Biomarkers for Atopic Dermatitis Mitochondrion by Learning Algorithms

Junhao Xu,1,* Xinyu Pan,2,* Miao Zhang,2,* Kairong Sun,2 Zihan Li,1 Juan Chen2 1The Second Clinical Medical College, Zhejiang Chinese Medical University, Hangzhou, Zhejiang, 310053, People’s Republic of China; 2College of Basic Medical Science, Zhejiang Chinese Medical University, Hangzhou, Zhejiang, 310053, People’s Republic of China*These authors contributed equally to this workCorrespondence: Juan Chen, Email [email protected]: Atopic dermatitis (AD) is a common inflammatory skin condition characterized by erythema and pruritus. Its precise pathogenesis remains unclear, though factors such as genetic predisposition, autoantigen response, allergen exposure, infections, and skin barrier dysfunction are involved. Research suggests a correlation between AD and mitochondrial dysfunction, as well as oxidative stress in skin tissues.Methods: Skin sample datasets related to AD (GSE36842, GSE120721, GSE16161, and GSE121212) were retrieved from the GEO database. Differential gene analysis identified differentially expressed genes (DEGs) in AD. Three potential biomarkers—COX17, ACOX2, and ADH1B—were identified using LASSO and Support Vector Machine (SVM) algorithms. These biomarkers were validated through ROC curve analysis, nomogram modeling, calibration curves, and real-time PCR. Immune infiltration analysis assessed correlations of the biomarkers. Additionally, single-cell analysis of the GSE153760 dataset identified nine cell clusters and confirmed expression patterns of the three hub genes.Results: Differential analysis identified 150 upregulated and 367 downregulated genes. Enrichment analysis revealed significant pathways related to mitochondrial function, oxidative stress, and energy metabolism in skin samples from AD patients. Area under the curve (AUC) values for biomarkers COX17, ACOX2, and ADH1B were 1.000, 0.928, and 0.895, respectively, indicating strong predictive capacity. qPCR results showed COX17 was highly expressed in AD lesions, while ACOX2 and ADH1B were higher in normal skin, consistent with previous findings. Correlation analysis indicated ACOX2 and ADH1B were positively correlated with resting mast cells but negatively with activated T cells and NK cells, while COX17 showed a positive correlation with activated T cells and a negative correlation with resting mast cells.Conclusion: This study suggests that the hub genes COX17, ACOX2, and ADH1B may serve as potential biomarkers in the pathogenesis of AD. These findings could provide insights for the treatment and prognosis of AD and related inflammatory skin conditions. Keywords: atopic dermatitis, mitochondria, learning algorithm, biomarkers, oxidative stress, immune infiltration