Psoriasis represents a multifaceted and debilitating immune-mediated systemic ailment afflicting millions globally. Despite the continuous discovery of biomarkers associated with psoriasis, identifying lysosomal biomarkers, pivotal as cellular metabolic hubs, remains elusive.
We employed a combination of differential expression analysis and weighted gene co-expression network analysis (WGCNA) to initially identify lysosomal genes. Subsequently, to mitigate overfitting and eliminate collinear genes, we applied 12 machine learning algorithms to screen robust lysosomal genes. These genes underwent further refinement through random forest (RF) and Lasso algorithms to ascertain the final hub lysosomal genes. To assess their predictive efficacy, we conducted receiver operating characteristic (ROC) analysis and verified the expression of diagnostic biomarkers at both bulk and single-cell levels. Furthermore, we utilized single-sample gene set enrichment analysis (ssGSEA), CIBERSORT, and Pearson’s correlation analysis to elucidate the association between immune phenotypes and hub lysosomal genes in psoriatic samples. Finally, employing the Cellchat algorithm, we explored potential mechanisms underlying the participation of these hub lysosomal genes in cell-cell communication.
Functional enrichment analyses revealed a close association between psoriasis and lysosomal functions. Subsequent intersection analysis identified 19 key lysosomal genes, derived from DEGs, phenotypic genes of WGCNA, and lysosomal gene sets. Following the exclusion of collinear genes, we identified 11 robust genes, further refined through RF and Lasso, yielding 3 hub lysosomal genes (S100A7, SERPINB13, and PLBD1) closely linked to disease occurrence, with high predictive capability for disease diagnosis. Concurrently, we validated their relative expression in separate bulk datasets and single-cell datasets. A nomogram based on these hub genes may offer clinical advantages for patients. Notably, these three hub genes facilitated patient classification into two subtypes, namely metabolic-immune subtype 1 and signaling subtype 2. CMap analysis suggested butein and arachidonic fasudil as preferred treatment agents for subtype 1 and subtype 2, respectively. Finally, through Cellchat and correlation analysis, we identified PRSS3-F2R as potentially promoting the expression of hub genes in the psoriasis group, thereby enhancing keratinocyte-fibroblast interaction, ultimately driving psoriasis occurrence and progression.
Our study identifies S100A7, SERPINB13, and PLBD1 as potential diagnostic biomarkers, offering promising prospects for more precisely tailored psoriatic immunotherapy designs.