Psoriasis is a chronic immune-mediated inflammatory skin disease characterized by dysregulated keratinocyte proliferation, immune cell infiltration, and systemic comorbidities. Despite the identification of numerous genetic susceptibility loci for psoriasis through genome-wide association studies (GWAS), their functional roles and underlying causal contributions to psoriasis pathogenesis remain largely unclear. Integrating multi-omics data with causal inference approaches, such as Mendelian randomization (MR), represents a promising strategy for addressing this gap and identifying key regulatory genes. We integrated transcriptome data from the Gene Expression Omnibus database (GEO) database (GSE14905 and GSE30999) and performed weighted gene co-expression network analysis and differential expression analysis to identify psoriasis-related genes. Protein-protein interaction networks and four centrality algorithms (Maximal Clique Centrality (MCC), Maximum Neighborhood Component (MNC), Edge Percolated Component (EPC), and Degree centrality) were applied to identify hub genes, and machine learning methods (least absolute shrinkage and selection operator regression, Random Forest, and artificial neural network) were used to screen diagnostic biomarkers. Immune infiltration analysis was performed using CIBERSORT, and the causal association of signature genes with psoriasis was examined using two-sample MR with single-cell expression quantitative trait locus data from the OneK1K cohort and GWAS summary statistics from FinnGen. Module genes that were significantly associated with psoriasis were identified in our study, among which 19 hub genes were screened. Using machine learning approaches, we further refined these findings to seven signature genes. The diagnostic model based on these seven genes achieved an area under the curve of 0.980. Immune infiltration analysis revealed strong associations between CCAAT/enhancer-binding protein delta (CEBPD) and activated CD4 + memory T cells and follicular helper T cells, and between zinc finger protein 36 (ZFP36) and M1 macrophages. MR analysis demonstrated that higher CEBPD expression in CD8 + S100B + T cells was protective (odds ratio [OR] = 0.795, P = 0.015), whereas higher expression of the ZFP36 gene in monocytes was a risk factor (OR = 1.214, P = 0.043) for psoriasis. Our study identified a robust seven-gene signature with high diagnostic accuracy for psoriasis and provided genetic evidence for the cell type-specific causal role of CEBPD and ZFP36. These findings enhance our understanding of psoriasis pathogenesis and suggest potential targets for developing cell-selective immunomodulatory therapies.
