Housekeeping gene dysregulation in psoriasis: integrative multi-cohort and single-cell analysis reveals keratinocyte-centric molecular mechanisms and diagnostic biomarkers.

BACKGROUND

Psoriasis is a chronic immune-mediated skin disease driven by the interleukin-23/interleukin-17 cytokine axis, yet its immunopathogenesis remains incompletely understood. Housekeeping genes, traditionally considered stably expressed across tissues and cell types, have not been systematically investigated for their role in psoriasis. Here, we aimed to identify psoriasis-associated housekeeping genes and explore their molecular mechanisms and clinical implications.

METHODS

We integrated multi-cohort data and identified psoriasis-associated housekeeping genes using weighted gene co-expression network analysis combined with differential expression analysis. Single-cell transcriptomic analysis was performed to identify cell-type specific expression patterns, while ligand-receptor interaction analysis was applied to evaluate pathway activation and interactions with downstream target genes. In addition, multiple diagnostic models were established for psoriasis detection.

RESULTS

We identified 34 housekeeping genes associated with psoriasis and observed that the co-expression relationships between six genes (APOL2, DCUN1D3, UBE2F, HIGD1A, PPIF, and STAT3) and known psoriasis-related genes differed significantly between diseased and healthy individuals. Furthermore, single-cell transcriptomic analysis revealed that these housekeeping genes were differentially expressed primarily in basal, spinous, supraspinous, and proliferating keratinocytes. Ligand-receptor interaction analysis demonstrated significant activation of the IL - 17, IL - 6, and midkine (MK) pathways within keratinocyte subpopulations, which led to the upregulation of STAT3, EIF5A, and RAN, thereby promoting keratinocyte hyperproliferation and enhancing immune reactivity. Finally, among the various diagnostic models developed, the averaged neural network (avNNet) model emerged as the best performer, achieving over 90% classification accuracy across multiple independent datasets. Moreover, its scores were strongly correlated with the Psoriasis Area and Severity Index (correlation coefficient = 0.74, P = 4.4e-47).

CONCLUSIONS

This study redefines housekeeping genes as dual-function regulators in psoriasis pathogenesis, with the avNNet model enabling clinical translation of these molecular insights toward precision-targeted therapies and biomarker-based management strategies.