Bioinformatics meets machine learning: identifying circulating biomarkers for vitiligo across blood and tissues.

Abstract

Background: Vitiligo is a skin disorder characterized by the progressive loss of pigmentation in the skin and mucous membranes. The exact aetiology and pathogenesis of vitiligo remain incompletely understood.

Methods: First, a microarray dataset of blood samples from multiple patients with vitiligo was collected from GEO database.The limma package was used to analyze the microarray data and identify significant differentially expressed genes (DEGs). The merged microarray data were then used for WGCNA to identify modules of features genes. DEGs selected with the limma package and module genes derived from the WGCNA were intersected using the Venn package in R. Enrichment analyses were performed on the overlapping genes, including Gene Ontology and Kyoto Encyclopedia of Genes and Genomes methodology. Advanced screening was performed using the least absolute shrinkage and selection operator and support vector machine techniques from the machine learning toolkit. CIBERSORT was used to analyse the immune cell composition in the microarray data to assess the relationships among these genes and immune cells. Biological samples were obtained from the patients, and gene expression analysis was performed to evaluate the levels of core genes throughout the progression of vitiligo. Finally, we obtained the microarray datasets GSE53146 and GSE75819 from the affected skin of vitiligo patients and GSE205155 from healthy skin to perform expression analysis and gene set enrichment analysis of the hub genes.

Results: Two hub genes, HMGA1 and PSMD13, were identified via machine learning and WGCNA. The analysis of immune cell infiltration suggested that different immune cell types could play a role in the progression of vitiligo. Moreover, these hub genes exhibited varying degrees of association with immune cell profiles. qRT-PCR analysis of blood samples from vitiligo patients revealed notable downregulation of the hub genes. Analysis of the microarray datasets derived from skin lesions revealed that HMGA1 expression levels remained relatively stable, whereas PSMD13 expression levels markedly decreased.

Conclusion: PSMD13 may influence vitiligo development via the Nod-like receptor signaling pathway and could serve as a potential diagnostic marker for evaluating skin lesions in vitiligo.