Machine learning-derived R-loop-related diagnostic signature for psoriasis: a multi-cohort and single-cell validation study
- 1Department of Dermatology, The First Affiliated Hospital of Nanjing Medical University
Abstract
Background: Psoriasis is a chronic inflammatory skin disease characterized by dysregulated immune responses and keratinocyte (KC) hyperproliferation. Although therapeutic strategies have advanced considerably, achieving durable long-term remission remains challenging. R-loops are three-stranded nucleic acid structures involved in genomic stability and inflammatory regulation in multiple diseases; however, their role in psoriasis remains poorly understood.
Objective: To investigate the involvement of R-loops in psoriasis, develop an R-loop-related diagnostic signature, and identify key molecular drivers of disease pathogenesis.
Methods: Transcriptomic data from five public bulk RNA expression cohorts (GSE30999, GSE41664, GSE13355, GSE6710, and GSE14905) and one single-cell RNA sequencing (scRNA-seq) dataset (GSE173706) were integrated for analysis of psoriatic and non-lesional skin samples. R-loop-related differentially expressed genes were identified and used to construct an R-loop-related diagnostic signature (RRDS) using the Elastic Net machine learning algorithm. Model performance was evaluated using three complementary analytical approaches, and SHapley Additive exPlanations (SHAP) analysis was performed to quantify the contribution of individual genes. scRNA-seq data were further analyzed to validate the findings at the cellular level. In addition, the biological function of SRPK1 was examined in HaCaT cells by siRNA-mediated knockdown, quantitative real-time PCR (qRT-PCR), and Cell Counting Kit-8 (CCK-8) assays.
Results: R-loop activity was significantly elevated in psoriatic lesions compared with non-lesional skin. We established and validated an eight-gene RRDS that accurately distinguished psoriatic from healthy skin. SHAP analysis identified SRPK1 as the most influential feature in the diagnostic model. Single-cell analysis further localized SRPK1 to a pathogenic KC_4 subpopulation at a critical stage of keratinocyte differentiation. Functional validation in HaCaT cells confirmed efficient SRPK1 knockdown by qRT-PCR and demonstrated that SRPK1 silencing significantly suppressed keratinocyte proliferation over 96 hours, as determined by CCK-8 assays (P < 0.001).
Conclusion: This study provides the first evidence linking R-loop dysregulation to psoriasis. We developed a R-loop-related diagnostic signature and identified SRPK1 as a potential key driver of psoriasis pathogenesis. Together with the in vitro findings, these results suggest that SRPK1 may serve as both a promising diagnostic biomarker and a potential therapeutic target in psoriasis.
Related
- Development and External Validation of an M2 Macrophage-Related Prognostic Signature in Breast Cancer Using Multi-Omics Data and Machine Learning.
- Identification of psoriasis-associated hub genes by integrated transcriptomics and machine learning
- A mucosal barrier injury related six gene transcriptomic signature associated with ulcerative colitis using machine learning and public transcriptomic data
- Characterization of the crotonylation landscape in gastric cancer: a machine learning-derived prognostic signature and the oncogenic role of TIMP1
- Integrated proteomic, genetic, and machine-learning analyses reveal a CXCL11-centered inflammatory signature associated with systemic activity in Sjögren’s disease
BibTeX
@article{feng2026machine,
title = {{Machine learning-derived R-loop-related diagnostic signature for psoriasis: a multi-cohort and single-cell validation study}},
author = {Yeqi Feng and Lebin Song and Xiaozheng Jiang and Hongyu Ruan and Jiayi Zhao and Yan Lu},
journal = {Frontiers in Immunology},
year = {2026},
month = oct,
doi = {10.3389/fimmu.2026.1839925},
url = {https://www.frontiersin.org/articles/10.3389/fimmu.2026.1839925/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Machine learning-derived R-loop-related diagnostic signature for psoriasis: a multi-cohort and single-cell validation study
AU - Yeqi Feng
AU - Lebin Song
AU - Xiaozheng Jiang
AU - Hongyu Ruan
AU - Jiayi Zhao
AU - Yan Lu
PY - 2026
DA - 2026/10/05
JO - Frontiers in Immunology
DO - 10.3389/fimmu.2026.1839925
UR - https://www.frontiersin.org/articles/10.3389/fimmu.2026.1839925/full
AB - Background: Psoriasis is a chronic inflammatory skin disease characterized by dysregulated immune responses and keratinocyte (KC) hyperproliferation. Although therapeutic strategies have advanced considerably, achieving durable long-term remission remains challenging. R-loops are three-stranded nucleic acid structures involved in genomic stability and inflammatory regulation in multiple diseases; however, their role in psoriasis remains poorly understood. Objective: To investigate the involvement of R-loops in psoriasis, develop an R-loop-related diagnostic signature, and identify key molecular drivers of disease pathogenesis. Methods: Transcriptomic data from five public bulk RNA expression cohorts (GSE30999, GSE41664, GSE13355, GSE6710, and GSE14905) and one single-cell RNA sequencing (scRNA-seq) dataset (GSE173706) were integrated for analysis of psoriatic and non-lesional skin samples. R-loop-related differentially expressed genes were identified and used to construct an R-loop-related diagnostic signature (RRDS) using the Elastic Net machine learning algorithm. Model performance was evaluated using three complementary analytical approaches, and SHapley Additive exPlanations (SHAP) analysis was performed to quantify the contribution of individual genes. scRNA-seq data were further analyzed to validate the findings at the cellular level. In addition, the biological function of SRPK1 was examined in HaCaT cells by siRNA-mediated knockdown, quantitative real-time PCR (qRT-PCR), and Cell Counting Kit-8 (CCK-8) assays. Results: R-loop activity was significantly elevated in psoriatic lesions compared with non-lesional skin. We established and validated an eight-gene RRDS that accurately distinguished psoriatic from healthy skin. SHAP analysis identified SRPK1 as the most influential feature in the diagnostic model. Single-cell analysis further localized SRPK1 to a pathogenic KC_4 subpopulation at a critical stage of keratinocyte differentiation. Functional validation in HaCaT cells confirmed efficient SRPK1 knockdown by qRT-PCR and demonstrated that SRPK1 silencing significantly suppressed keratinocyte proliferation over 96 hours, as determined by CCK-8 assays (P < 0.001). Conclusion: This study provides the first evidence linking R-loop dysregulation to psoriasis. We developed a R-loop-related diagnostic signature and identified SRPK1 as a potential key driver of psoriasis pathogenesis. Together with the in vitro findings, these results suggest that SRPK1 may serve as both a promising diagnostic biomarker and a potential therapeutic target in psoriasis.
SN - 1664-3224
ER -