Open Access

ARTICLE

Integrative transcriptomic analysis identifies resistin as a candidate serum marker associated with atopic dermatitis severity

Shengjie Xue^#, Shiyu Ni^#, Zehao Lan, Jieyue Liao^*

Department of Dermatology, Second Xiangya Hospital, Hunan Key Laboratory of Medical Epigenomics, Clinical Medical Research Center of Major Skin Diseases and Skin Health of Hunan Province, Central South University, Changsha, China

* Corresponding Author: Jieyue Liao. Email:
# These authors contributed equally to this work

(This article belongs to the Special Issue: Inflammation in Disease: When Cytokine Conversations Tell the Whole Story)

European Cytokine Network 2026, 37(3), 331-348. https://doi.org/10.32604/ecn.2026.086314

Received 28 May 2026; Accepted 14 September 2026; Issue published 28 September 2026

Abstract

Background: Atopic dermatitis (AD) is associated with systemic immune dysregulation, but peripheral molecular markers that reliably reflect clinical severity remain insufficiently validated. Using complementary bulk and single-cell transcriptomics, we prioritized peripheral markers and evaluated them in independent clinical cohorts. Methods: Public peripheral blood mononuclear cell (PBMC) transcriptomes were analyzed by differential-expression analysis, weighted gene co-expression network analysis (WGCNA) with power-sensitivity and preservation assessments, immune-cell deconvolution, LASSO, and random forest. Descriptive single-gene receiver operating characteristic (ROC) curves summarized candidate discrimination in the discovery cohort. Selected transcripts were evaluated by RT-qPCR in a PBMC cohort, and serum resistin by ELISA in a separate non-overlapping cohort. Severity associations were examined using multivariable models adjusted for age, sex, and BMI. Results: After correction for multiple module–trait comparisons, WGCNA identified an internally preserved conventional module inversely associated with AD. Power-stable WGCNA-unassigned genes formed an exploratory pool intersected with AD-associated differentially expressed genes. LASSO retained RETN, AZU1, RNASE3, and MS4A3 as computational candidates, and all four genes received relatively high importance rankings in the complementary random-forest analysis. Descriptive within-cohort analyses yielded apparent areas under the curve of 0.750, 0.736, 0.706, and 0.689, respectively, and were not interpreted as independent diagnostic validation. RT-qPCR showed higher RETN, AZU1, and RNASE3 expression in AD PBMC preparations. Deconvolution and single-cell analysis supported a predominantly classical-monocyte-associated expression context for RETN. Serum resistin concentrations were higher in AD and remained positively associated with SCORAD and DLQI after adjustment, although bootstrap analyses indicated uncertainty. Conclusions: Resistin is a candidate peripheral marker associated with AD severity. Its mechanistic significance and clinical utility require functional studies and independent prospective longitudinal validation.

---

Keywords

---

Supplementary Material

Supplementary Material File

---