Identification of psoriasis-associated hub genes by integrated transcriptomics and machine learning

This study aims to systematically identify key hub genes in psoriasis pathogenesis through the integration of bioinformatics, machine learning and experimental validation methods. Four psoriasis microarray datasets were retrieved from the GEO database, with differentially expressed genes identified through comparative analysis. A dual machine learning approach screened core hub genes, whose expression levels and diagnostic efficacy were validated in an independent cohort. CIBERSORT was employed for immune cell infiltration analysis, whilst a multi-database approach constructed an mRNA-miRNA-lncRNA regulatory network. Molecular docking technology assessed interactions between the environmental pollutant bisphenol A, predicted drugs, and hub genes. Finally, an in vitro psoriasis model was established by stimulating HaCaT cells with the M5 cytokine mixture, with hub gene expression validated via RT-qPCR. Results show that 77 co-differentially expressed genes were identified in psoriasis. Machine learning screened five key hub genes: CARHSP1, HPSE, PCLAF, SERPINB4, and SPRR2C. These were significantly upregulated in the validation cohort and demonstrated strong discriminative value for distinguishing psoriatic lesional skin from paired non-lesional skin, rather than serving as standalone diagnostic biomarkers for psoriasis. Immune cell infiltration analysis revealed extensive associations with immune subpopulations including T cells and macrophages. Molecular docking computationally predicted a potential binding interaction between the environmental pollutant BPA and the hub gene HPSE, which awaits future experimental validation. Concurrently, Genistein and HPSE were suggested a potential interaction that awaits functional validation. In vitro experiments confirmed significant upregulation of HPSE, PCLAF, and SPRR2C mRNA levels in M5-induced psoriasis cell models. HPSE emerged as a multifunctional core gene associated with immune dysregulation. Molecular docking additionally predicted a potential interaction between the environmental pollutant BPA and HPSE; however, this interaction remains a computational hypothesis requiring experimental validation.