Related Experiment Videos
Integrated Single-Cell Transcriptomics, Multi-Omics, and Mendelian Randomization Reveal Ferroptosis-Related Molecular
Mengyue Xu1, Yongjie Zuo2, Hongxia Liu2
1Department of Dermatology, The First Affiliated Hospital of Henan University of Chinese Medicine, Henan Province, People's Republic of China.
Background:
Ferroptosis contributes to inflammatory skin diseases, but its role in psoriasis remains incompletely understood. We performed an integrated multi‑omics analysis combining single‑cell transcriptomics, bulk transcriptomics, machine learning, consensus clustering, and Mendelian randomization (MR) to characterize ferroptosis‑related molecular features, intercellular communication, immune microenvironment remodeling, molecular subtypes, and causal biomarkers in psoriasis.
Methods:
Single‑cell RNA‑seq data (GSE151177) were analyzed to quantify ferroptosis activity across cell populations and infer intercellular communication networks. Bulk transcriptomic datasets (GSE30999, GSE78097) were integrated for differential expression and machine‑learning feature selection with SHAP interpretation. Consensus clustering, immune infiltration, functional enrichment, and drug repurposing analyses were performed. Cis‑eQTLs from eQTLGen (n=31,684) served as instruments for two‑sample MR with psoriasis GWAS from FinnGen (13,667 cases, 395,092 controls).
Results:
Macrophages/monocytes and fibroblasts showed the highest ferroptosis scores. MIF signaling (MIF-CD74/CD44, MIF-CD74/CXCR4, MIF-ACKR3) dominated keratinocyte outgoing communication. Transcriptomic integration yielded 72 overlapping ferroptosis-related genes; machine learning identified eight feature genes (CD24, EFNB2, DDX39A, NDRG4, GPRC5A, S100A8, PRSS8, ITPKC). CD24 (AUC=0.979) and S100A8 (AUC=0.974) exhibited top diagnostic performance. Consensus clustering stratified patients into two subtypes (C1, C2); all eight genes were highly expressed in C1, while C2 showed upregulated phagocytic but downregulated innate immune pathways. Drug enrichment suggested MS-275, genistein, and β-adrenoceptor agonists as potential ferroptosis-modulating compounds. MR demonstrated that only PRSS8 eQTL had a nominally significant causal association with psoriasis risk (P < 0.05, β = 0.0073).
Conclusion:
Our multi‑omics study delineates the ferroptosis‑related landscape in psoriasis, identifies MIF signaling as a key keratinocyte communication axis, establishes a high‑performance diagnostic model with eight feature genes, defines two distinct immune‑metabolic subtypes, and provides genetic evidence supporting PRSS8 as a putative causal biomarker for psoriasis susceptibility, offering novel insights and potential ferroptosis‑based biomarkers and therapeutic targets for psoriasis.