Skip to main content
NCBI home page
Clinical, Cosmetic and Investigational Dermatology logoLink to Clinical, Cosmetic and Investigational Dermatology
. 2026 Sep 21;19:636697. doi: 10.2147/CCID.S636697

Integrative Transcriptomic and Genetic Analyses Identify CDC20 as a Potential Biomarker Associated with Cell Cycle and Immune-Inflammatory Processes in Psoriasis

Lingyi Yang 1,*, Wenshen Mo 1,*, Gang Wang 2, Xiaowei Wu 3,✉
PMCID: PMC13614442  PMID: 42799260

Abstract

Objective

To identify psoriasis-associated candidate genes and regulatory features via integrated transcriptomic, immune and genetic analyses, and validate them in an imiquimod-induced psoriasis-like murine model.

Methods

Public GEO transcriptomic datasets were integrated after batch-effect correction. We performed differential expression, WGCNA, machine-learning-based feature selection and diagnostic modelling, CIBERSORT immune infiltration estimation, transcription-factor regulation, drug enrichment, colocalization and SMR analyses. In-vivo histological and Western blot validation was conducted in mice.

Results

We obtained 767 candidate genes enriched in cell-cycle and immune-inflammatory pathways. The Lasso-XGBoost model (AUC = 0.923) yielded 11 candidate genes that well separate psoriasis patients from healthy controls. Drug enrichment identified Lucanthone. SMR revealed consistent expression-genetic directionality for LYN and IL1RN. Notably, the 11 machine-learning-derived signature genes mainly mark psoriasis disease activity and differ from the SMR-identified genetically supported causal genes LYN and IL1RN. Increased CDC20, CCNB1 and CDK1 protein levels were confirmed in mouse lesional skin.

Conclusion

This study suggests that CDC20 may serve as a potential biomarker for cell cycle and immune-inflammatory processes in psoriasis, providing a basis for further mechanistic and translational studies.

Keywords: WGCNA, machine learning, immune cell infiltration, colocalization, mendelian randomization

Introduction

Psoriasis is a common chronic, relapsing, immune-mediated inflammatory skin disease characterized by excessive proliferation and abnormal differentiation of epidermal keratinocytes, along with dermal immune cell infiltration, substantially impairing physical health and quality of life.1,2 Zou et al, based on Global Burden of Disease (GBD) data, reported an increasing burden of psoriasis in China, highlighting its growing public health significance.3 In addition, Rendon and Schäkel4 emphasized that psoriasis imposes a substantial disease burden through its chronic and recurrent course and its adverse effects on patients’ physical and psychological well-being and quality of life. It is widely accepted that its pathogenesis involves genetic susceptibility, immune dysregulation, and environmental factors.5

Despite substantial progress, several limitations remain in current psoriasis research. Zhou et al6 applied LASSO and SVM-RFE combined with immune infiltration analysis to identify UGGT1, CCNE1, MMP9, and ARHGEF28 as potential diagnostic biomarkers for psoriasis and further validated their expression using RT-qPCR and immunohistochemistry; however, their study mainly focused on biomarker identification and immune-cell associations, without integrating genetic causal inference or systematic transcriptional regulatory analyses. Deng et al7 employed a tree-based machine learning approach to characterize gene regulatory networks in psoriasis and identified key regulators and interferon signaling associated with psoriatic inflammation; nevertheless, genetic causal inference was not incorporated to further evaluate the causal relevance of the identified regulators. Strange et al8 conducted a genome-wide association study and identified several novel psoriasis susceptibility loci involving immune-related genes, including IL28RA, REL, IFIH1, ERAP1, TRAF3IP2, NFKBIA, and TYK2; however, genetic association alone cannot fully establish the functional or causal roles of the implicated genes. More recently, Si et al9 integrated proteome-wide Mendelian randomization, transcriptomic analyses, colocalization, and other multi-omics approaches to prioritize potential therapeutic targets for psoriasis; however, their study primarily focused on therapeutic target discovery rather than the development of an integrated diagnostic framework incorporating transcriptomic feature selection, machine learning, immune infiltration, and transcriptional regulation. Collectively, these studies highlight the need for an integrative framework combining transcriptomic analysis, robust machine learning, immune and regulatory network characterization, genetic causal inference, and therapeutic target exploration.

Recent advances in high-throughput sequencing and bioinformatics have established a framework for systematic interrogation of psoriasis at the transcriptomic level. Integrative strategies that combine multi-omics, network-based analyses and machine learning enable the dissection of disease mechanisms across layers of expression, regulation, immunity and genetics, thereby enhancing the robustness of gene prioritisation and providing a more rigorous basis for precision diagnosis and targeted intervention.5,10

Building on this framework, we combined differential expression analysis with weighted gene co-expression network analysis and applied a Lasso–XGBoost ensemble to improve the robustness of gene prioritisation. Immune infiltration profiling, transcriptional regulatory network construction and drug enrichment analyses were integrated to delineate the molecular landscape of psoriasis within the context of the immune microenvironment and its regulatory architecture. Colocalization and SMR analyses were then undertaken to support the genetic causality of candidate genes. Through this multi-layered strategy, we aimed to identify key genes and therapeutic targets with diagnostic and translational relevance, thereby providing a framework for precision diagnosis and personalised treatment in psoriasis. Experimentally, an imiquimod-induced murine model of psoriasis was used to assess histopathological changes by H&E staining and to validate the protein expression levels of CDC20, CCNB1, and CDK1 by Western blotting.

Materials and Methods

Psoriasis gene expression datasets were retrieved from the GEO database. After normalisation and correction for batch effects, differentially expressed genes (DEGs) were identified. Weighted gene co-expression network analysis (WGCNA) was then applied to construct co-expression networks and to define disease-associated core modules, which were intersected with DEGs to yield candidate genes of interest.

Twelve feature selection algorithms were subsequently implemented, generating 120 integrated machine learning models for gene prioritisation and diagnostic modelling. Immune cell infiltration analysis was undertaken to characterise the immune microenvironment, alongside the construction of transcription factor regulatory and drug–gene interaction networks. Colocalization and SMR analyses were further employed to delineate the genomic context of key genes and to support their causal association with psoriasis phenotypes. The overall workflow is outlined in Figure 1.

Figure 1.

A diagram showing psoriasis pathogenesis with immune cells and therapeutic targets. The diagram outlines psoriasis pathogenesis, highlighting pro-inflammatory immune cells and CDC20′s role in keratinocyte cell-cycle dysregulation. Immune cells lead to inflammation, affecting CDC20 in keratinocytes, linked to gene overexpression and cell cycle issues. Lucanthone is depicted as a therapy targeting CDC20. The cell cycle is shown with phases G2, M and S. MELK and CDK1 connect to CDC20, while LYN and IL1RN relate to SMR genetic association. Skin lesions illustrate the psoriasis phenotype. The diagram suggests Lucanthone′s potential in targeting CDC20 to alleviate psoriasis symptoms.

Open in a new tab

Schematic of psoriasis pathogenesis. Pro‑inflammatory immune cells trigger CDC20‑dependent keratinocyte cell‑cycle dysregulation to drive psoriatic lesions. Lucanthone serves as a potential therapeutic inhibitor. Green arrows denote promotion, and red‑| bars denote inhibition.

Data Sources and Preprocessing

Transcriptomic datasets for psoriasis were obtained from the Gene Expression Omnibus (GEO, https://www.ncbi.nlm.nih.gov/geo). GSE13355, GSE6710 and GSE14905 were used as the training set, while GSE30999, GSE106087 and GSE166388 served as independent validation cohorts. Batch effects were corrected using surrogate variable analysis (SVA) in combination with the ComBat algorithm, and principal component analysis (PCA) was applied to assess normalisation and batch correction, ensuring comparability across datasets.

For colocalization analysis, eQTL data for core genes were retrieved from the GWAS database (https://gwas.mrcieu.ac.uk/), and psoriasis GWAS summary statistics were obtained from the FinnGen consortium (https://www.finngen.fi/en/access_results).

Identification of Differentially Expressed Genes and Construction of the Co-Expression Network

Differential expression analysis was performed on the training set using the limma package, with thresholds set at |log2FC| > 0.585 (1.5-fold change) and adjusted P < 0.05 to define DEGs between psoriasis and normal tissues. Results were visualised using volcano plots and heatmaps generated with the ggplot2 package.

Gene co-expression networks were constructed using the WGCNA package, followed by phenotype association analysis. Genes with low variability (standard deviation ≤ 0.5) were excluded. A soft-thresholding power was selected to achieve a scale-free topology fit index of R2 ≥ 0.8. The minimum module size was set to 60 genes, and modules with a correlation coefficient ≥ 0.75 were merged.

Functional and Pathway Enrichment Analysis

Intersecting genes derived from DEGs and WGCNA core modules were subjected to Gene Ontology (GO) annotation and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis using the clusterProfiler package within the Bioconductor framework. Significantly enriched terms were defined at P < 0.05 and visualised using bubble plots, bar charts and pathway network diagrams.

The intersecting genes were then uploaded to the STRING database (restricted to Homo sapiens) with a confidence score ≥ 0.4 to obtain protein–protein interaction (PPI) data. The resulting PPI network was imported into Cytoscape (version 3.10.4) for visualisation, and the top 20 ranked genes were selected as key candidates.

Machine Learning Model Construction and Key Gene Selection

Based on the training transcriptomic data, 12 classical algorithms, including Lasso, XGBoost and random forest (RF), were incorporated and combined to generate 120 classification models. Ten-fold cross-validation was implemented for Lasso via cv.glmnet within the training set to select the optimal λ (lambda.min). For XGBoost, five-fold internal cross-validation was applied to determine the optimal number of boosting rounds (nround), with fixed hyper-parameters: max.depth = 2, eta = 1. For random forest, ntree = 1000 and nodesize = 5 were used. For GBM, ten-fold cross-validation was performed, with fixed hyper-parameters: interaction.depth = 3, n.minobsinnode = 10, shrinkage = 0.001. Model performance was evaluated using the area under the curve (AUC), accuracy and F1 score, and the optimal model was selected accordingly.

Feature genes derived from the optimal model were ranked according to statistical significance (P values), and the top 11 genes were identified as key candidates. Their classification performance was further validated in independent external datasets, with discriminatory efficacy in both the training and validation cohorts assessed by AUC.

Immune Cell Infiltration and Correlation Analysis of Key Genes

Immune cell infiltration was estimated using the CIBERSORT deconvolution algorithm. Visualisation and statistical analyses were performed using the reshape2, ggpubr and corrplot packages. Stacked bar plots were generated to display the relative proportions of immune cells in control and disease groups, while intergroup differences were assessed using boxplots and the Wilcoxon rank-sum test. Spearman correlation analysis based on the disease cohort was used to construct an immune cell interaction heatmap, thereby delineating co-infiltration patterns.

To further characterise the relationship between key genes and the immune microenvironment, Spearman correlation analysis was conducted by integrating gene expression profiles with immune infiltration data. Data processing and statistical testing were performed using limma, reshape2, ggpubr and ggExtra, with correlation coefficients and P values calculated accordingly. The results were visualised using lollipop plots, in which node size represented correlation strength and node colour denoted statistical significance, providing a systematic view of the association between key genes and immune cell infiltration.

Regulatory Network Construction and Drug Enrichment Analysis

Human transcription factor–target gene regulatory interactions were obtained from the TRRUST database. Core genes identified from the optimal model were used as input to construct a transcription factor–gene regulatory network using Cytoscape (https://cytoscape.org/).

Drug–gene interaction data were retrieved from the DSigDB database, and drug enrichment analysis was performed on the core gene set using the clusterProfiler package in R. Significantly enriched candidate drugs were identified using thresholds of P < 0.05 and adjusted P < 0.05. Enrichment results were visualised using bar plots and bubble plots generated with ggplot2, illustrating both statistical significance and gene enrichment patterns.

Genetic Colocalization Analysis

Genetic colocalization analysis between eQTL signals of core genes and psoriasis GWAS data was performed using the coloc package to assess the sharing of causal variants. Prior to analysis, SNP matching, quality control and locus filtering were conducted for both eQTL and GWAS datasets, with a ±100 kb window around the target gene defined as the region of interest. Posterior probabilities (PP) were estimated using the coloc.abf algorithm, with PP.H4 > 0.5 indicating significant colocalization between eQTL and disease-associated signals.

Colocalization results were further visualised using the locuscomparer package, illustrating association strength and local linkage disequilibrium structure.

Mendelian Randomization (SMR) Analysis

Psoriasis GWAS summary statistics were obtained from the FinnGen R12 release. Data quality control and harmonisation were performed in R, extracting key information including SNPs, alleles, effect sizes, standard errors and P values. SMR results were filtered using thresholds of PSMR < 0.05 for significance and PHEIDI > 0.05 to exclude horizontal pleiotropy, thereby identifying robust causal genes.

Results were visualised using the CMplot package, with Manhattan and circular plots displaying genome-wide association signals as −log10(P), and significant genes highlighted to illustrate the distribution of association signals across the genome.

Integrated Candidate Gene Selection and Visualisation

Differentially expressed genes, core genes derived from the optimal model, and SMR-identified causal genes were integrated, and their overlap was visualised using the ggvenn package in R. Gene sets were stratified into upregulated and downregulated groups according to logFC direction and the sign of the SMR effect estimate (bSMR). Venn diagrams were used to delineate the overlap between differential expression and causal association, and robust candidate genes were defined as those showing concordant expression patterns and causal effects.

Further visualisation of candidate genes was performed to characterise their genetic architecture. Gene-centric plots were generated using SMRLocusPlot to illustrate colocalization patterns between eQTL and GWAS signals, while SMREffectPlot was used to depict the magnitude and direction of causal effects, enabling systematic interpretation of the genetic regulatory mechanisms underlying candidate genes.

Establishment of the Psoriasis-Like Murine Model

Animal experiments were conducted using male BALB/c mice (6–8 weeks old, 30–35 g), obtained from Jiangsu Qinglongshan Biotechnology Co., Ltd. (Nanjing, China). Psoriasis-like skin inflammation was induced by topical application of 62.5 mg of 5% imiquimod (IMQ) cream (Tianfang Pharmaceutical Co., Ltd., Suzhou, China), as previously described.11 Mice were randomly assigned to two groups (n = 3 mice per group): a control group (vehicle, petrolatum) and a psoriasis-like model group (IMQ). The intervention period lasted 1 week.

Body weight, affected skin area and Psoriasis Area and Severity Index (PASI) scores were recorded every other day. A modified PASI scoring system was used to evaluate disease severity, with erythema, induration and scaling each graded on a scale from 0 (none) to 4 (most severe).12

At the end of the intervention, mice were humanely euthanized. Mice were deeply anesthetized by intraperitoneal injection of 20% urethane (1.5 g/kg), followed by immediate cervical dislocation to ensure rapid, complete, and irreversible euthanasia. Death was confirmed by the permanent cessation of spontaneous respiration and heartbeat. Lesional skin samples were then carefully collected for subsequent histological and molecular analyses, including hematoxylin and eosin (H&E) staining and Western blot (WB) analysis. All animal experimental procedures were approved by the Clinical Research Ethics Committee of the Affiliated Hospital of Qinghai University (Approval No. FYXC-2024-13). All experiments were conducted in accordance with the ARRIVE guidelines and the national regulations for the care and use of laboratory animals in China. Every effort was made throughout the study to minimize animal suffering and reduce the number of animals used.

Histological Analysis

Skin samples were fixed in 4% paraformaldehyde for 24 h, embedded in paraffin, and sectioned at 4 μm thickness. Sections were stained with haematoxylin and eosin (H&E) to assess histopathological changes, and epidermal thickness was measured accordingly.

Western Blot Analysis

Total protein was extracted from murine skin tissues, and protein concentration was determined using the BCA assay. Proteins were separated by SDS–PAGE and transferred onto PVDF membranes, followed by blocking with 5% non-fat milk for 1 h at room temperature. Membranes were incubated overnight at 4 °C with primary antibodies against CDC20 (UpingBio, YP-Ab-16676), CCNB1 (UpingBio, YP-Ab-16722), CDK1 (UpingBio, YP-Ab-16696) and β-actin (Proteintech, Cat. No. 66009-1-Ig), followed by incubation with appropriate secondary antibodies for 1 h at room temperature. Signals were detected using enhanced chemiluminescence (ECL), and relative protein expression levels were quantified using ImageJ, normalised to β-actin. Western blot analyses were performed using independent skin samples from three mice per group, with each mouse serving as an independent biological replicate (n = 3 biological replicates per group).

Results

Data Preprocessing, Differential Expression Analysis and WGCNA Module Identification

Training datasets were integrated and normalised prior to analysis. Principal component analysis (PCA) revealed pronounced batch effects before correction, with dispersed sample distribution (Figure 2A), whereas post-normalisation samples exhibited tight clustering, indicating effective removal of batch effects (Figure 2B).

Figure 2.

Five panels show PCA, volcano plot, module to trait analysis and heatmap of gene expression. The image A shows a PCA plot before batch correction with dispersed samples labeled as GSE13355, GSE14905 and GSE6710. The image B shows a PCA plot after correction with improved clustering of the same sample types. The image C shows a volcano plot of DEGs with logFC on the x-axis and negative log10 of p-value on the y-axis, highlighting significant up and downregulated genes. The image D shows a module to trait analysis with a heatmap identifying MEbrown as the key module with correlation values. The image E shows a heatmap of core gene expression differences across control and psoriasis samples, with projects GSE13355, GSE14905 and GSE6710 indicated and a color scale for expression levels.

Open in a new tab

Data preprocessing, differential expression, and WGCNA analysis. (A) PCA before batch correction showing batch effects. (B) PCA after correction showing improved clustering. (C) Volcano plot of DEGs. (D) Module–trait analysis identifying MEbrown as the key module. (E) Heatmap of core gene expression differences.

Differential expression analysis identified 950 genes associated with psoriasis, including 544 upregulated and 406 downregulated transcripts. Global expression patterns were visualised by volcano plots and heatmaps (Figure 2C and E).

Weighted gene co-expression network analysis (WGCNA) identified modules associated with disease status, among which the MEbrown module showed the strongest correlation with psoriasis (Figure 2D), being positively associated with disease (r = 0.77, P = 3 × 10−58) and negatively with controls (r = −0.77, P = 3 × 10−58), and was therefore defined as the core module.

An imiquimod-induced murine model recapitulated key features of psoriasis. Progressive erythema, epidermal thickening and scaling developed over time (Figure 3A and B). Histological analysis revealed that compared with the normal control group, psoriatic lesions exhibited significant epidermal hyperplasia and inflammatory cell infiltration. Meanwhile, the epidermal thickness, dermal thickness, and full-thickness skin thickness in the psoriasis group were increased compared with the control group (P < 0.05). (Figure 3C and D), confirming the robustness of the model.

Figure 3.

Images of a psoriasis mouse model showing skin changes and histological analysis. A diagram of a mouse with psoriasis lesions from day 2 to 7. Circular images show skin changes over these days, with psoriasis and normal labels. Histological images display H&E staining at 10 times and 4 times magnification, comparing normal and psoriatic skin. Bar graphs depict epidermal, dermal and skin thickness, with significant increases in the psoriasis group compared to normal, marked by asterisks indicating statistical significance.

Open in a new tab

IMQ-induced psoriasis mouse model and pathological validation. (A) Representative skin lesions showing erythema, thickening, and scaling from day 2–7. (B) Gross skin appearance of control and model groups. (C) H&E staining (4×, 10×) showing epidermal hyperplasia and inflammatory infiltration. (D) Quantitative analysis of epidermal, dermal, and full-thickness skin thickness in psoriasis and normal groups. **P < 0.01.

Key Gene Prioritization and Functional Enrichment Analysis

To identify genes closely associated with psoriasis, DEGs were intersected with genes from the WGCNA-derived core module (Figure 4), yielding 767 overlapping genes, representing 54.1% of the total.

Figure 4.

A 2-set Venn diagram showing overlap between DEG and WGCNA genes.

Open in a new tab

Overlap between DEGs and WGCNA core module genes. Venn diagram showing 767 intersecting genes selected as candidate genes for further analysis.

Functional enrichment analysis of these intersecting genes revealed biologically relevant patterns (Figure 5). Gene Ontology (GO) analysis (Figure 5A) indicated enrichment in biological processes including mitosis, nuclear division, organelle fission, and positive regulation of cytokine production. Cellular components were primarily associated with the spindle apparatus and centromeric regions of chromosomes, while molecular functions were enriched for chemokine activity, cytokine activity, and receptor binding. KEGG pathway analysis (Figure 5B) highlighted significant enrichment in cytokine–cytokine receptor interaction, keratinization, lipid metabolism, and atherosclerosis-related inflammatory and immune pathways.

Figure 5.

Two graphs: GO bar graph on gene enrichment and KEGG bubble graph on pathway enrichment. Panel A shows a bar graph of Gene Ontology enrichment analysis. The x-axis represents the count of genes and the y-axis lists terms grouped into Biological Process, Cellular Component and Molecular Function. Bars are color-coded by p-value. Key terms include mitotic nuclear division, spindle and chemokine activity. Panel B displays a KEGG pathway bubble graph. The x-axis is GeneRatio and the y-axis lists pathways. Bubble size indicates count and color represents p-value. Notable pathways include cytokine-cytokine receptor interaction and coronavirus disease. Both graphs highlight significant gene and pathway enrichments, with lower p-values indicating higher significance.

Open in a new tab

GO and KEGG enrichment analysis of candidate genes. (A) GO analysis showing enrichment in mitosis, nuclear division, cytokine regulation, and chemokine activity. (B) KEGG analysis showing enrichment in cytokine–receptor interaction, cell cycle, IL-17 signalling, and immune-related pathways.

Protein–protein interaction (PPI) network construction (Figure 6) revealed a densely connected network architecture. Network topology analysis identified CCNA2, CCNB1, CDC20, CDK1, and BUB1B as hub genes with high connectivity (Figure 7), suggesting their central role in the regulatory network underlying psoriasis pathogenesis.

Figure 6.

A diagram showing a protein protein interaction network graph highlighting hub genes CCNA2, CCNB1, CDC20 and CDK1. A network graph shows protein interactions with nodes labeled by gene names connected by lines. The dense core module in the center has many overlapping connections and radiates lines to surrounding nodes and smaller modules. Peripheral subnetworks form distinct clusters with internal connections, located in the upper right, upper middle, lower left and lower right, each linked to the core. Hub genes CCNA2, CCNB1, CDC20 and CDK1 are in the core, connected to many nodes. Lines represent protein interaction links, with no axis, scale, units, or edge weights shown.

Open in a new tab

PPI network and identification of hub genes. The network shows interactions among candidate genes, with CCNA2, CCNB1, CDC20, and CDK1 identified as key hubs.

Figure 7.

Hub genes CCNA2, CCNB1, CDC20, CDK1, BUB1B interact in psoriasis protein network. The image depicts a protein to protein interaction network illustrating interactions among key hub genes involved in psoriasis. The network is densely connected, with central hub genes labeled as CCNA2, CCNB1, CDC20, CDK1 and BUB1B, highlighted in red. Surrounding these are additional genes such as NCAPG, NUSAP1, KIF4A, MELK, BIRC5, UBE2C, TPX2, DLGAP5, CEP55, KIF2C, KIF11, TOP2A, KIF20A, CDCA8 and ASPM, shown in yellow and orange. Lines between the genes represent interactions, indicating a complex network architecture underlying psoriasis pathogenesis.

Open in a new tab

PPI network of core genes. Illustrates interactions among key hub genes in psoriasis.

Machine Learning Model Selection and Diagnostic Evaluation of Key Genes

Multiple machine learning algorithms were applied to the gene expression data to construct predictive models and evaluate their performance across datasets (Figure 8A). The Lasso+XGBoost model achieved the highest overall performance, with an AUC of 0.923. Other models, including Lasso+SVM and SVM, also demonstrated high predictive accuracy, with AUCs of 0.918 and 0.913, respectively. All models exhibited consistent performance across training and validation cohorts.

Figure 8.

Four panels: AUC heatmap, gene expression boxplots, ROC curves and Western blot analysis for psoriasis study. The image A shows a heatmap of AUC values across different datasets and training sets for various machine learning models. Models include Lasso+XGBoost, Lasso+SVM and others, with AUC values ranging from 0.857 to 0.923. Cohorts are labeled as GSE106087, GSE166388, GSE39999 and Train. The image B shows boxplots of gene expression levels for ASPH, BIRC5, CCNA2, CCNB1, CDC20, CDK1, CEP55, KIF11, MELK, NUSAP1 and TOP2A in control and psoriasis samples, with significance indicated as *P < 0.001. The image C shows ROC curves for psoriasis diagnosis, with sensitivity plotted against 1-specificity for genes like ASPH, BIRC5, CCNA2 and others, with AUC values listed. The image D shows Western blot analysis of protein expression for CCNB1, CDC20, CDK1 and beta-actin in healthy and psoriasis-like mouse skin, with molecular weights of 60kd, 50kd, 34kd and 43kd respectively.

Open in a new tab

Validation and diagnostic performance of key genes. (A) Heatmap of AUC values across datasets and training set. (B) Boxplots showing the expression levels of key genes in psoriasis and control samples. ***P < 0.001. (C) ROC curves for psoriasis diagnosis. (D) Western blot analysis of CDC20, CCNB1, and CDK1 protein expression in control and IMQ-induced psoriasis-like mouse skin. n = 3 biological replicates per group.

Based on AUC evaluation, the Lasso+XGBoost model was selected as optimal, from which 11 key feature genes were identified. Expression analysis revealed that these genes were significantly upregulated in psoriasis samples compared with controls (P < 0.001) (Figure 8B).

To evaluate the discriminatory performance of each key gene, ROC curve analysis was performed. All candidate genes exhibited relatively high AUC values in distinguishing psoriasis patients from healthy controls within the current dataset, with AUCs exceeding 0.88. CDC20 exhibited the highest diagnostic performance (AUC = 0.912), while CCNA2 (AUC = 0.902), CDK1 (AUC = 0.906) and CCNB1 (AUC = 0.901) also showed robust AUC values (Figure 8C). CDC20, CDK1, and CCNB1 were subsequently subjected to protein-level expression validation by Western blotting in the imiquimod-induced psoriasis-like mouse model. Protein levels of all three molecules were markedly increased in the psoriasis-like group compared with the control group, consistent with the transcriptomic expression trends (Figure 8D).

Immune Cell Infiltration Characteristics and Immune Correlation of Key Genes

Immune infiltration analysis was performed to quantitatively compare the proportions of 22 immune cell types between psoriasis and control samples. Significant differences in the infiltration abundance of multiple immune cell subsets were observed between the two groups (Figure 9A). Specifically, the psoriasis group exhibited markedly increased infiltration of CD4 memory-activated T cells, follicular helper T cells, regulatory T cells (Tregs), γδ T cells, resting NK cells, M0/M1 macrophages, activated dendritic cells, and neutrophils, whereas plasma cells, activated NK cells, resting dendritic cells, and resting mast cells were more abundant in controls. Further correlation analysis revealed that the 11 key genes were strongly positively correlated with activated dendritic cells and strongly negatively correlated with resting mast cells (Figure 9B).

Figure 9.

A box plot and a lollipop plot comparing immune cell fractions and CDC20 correlations in control and psoriasis. A two graph figure. A) A box plot comparing immune cell infiltration between Type, Control and psoriasis. X-axis label: immune cell types. Y-axis label: Fraction, unit not shown, range 0.0 to 0.6 with ticks at 0.2, 0.4, 0.6. Cell types shown include B cells naive, B cells memory, Plasma cells, T cells CD8, T cells CD4 naive, T cells CD4 memory resting, T cells CD4 memory activated, T cells follicular helper, T cells regulatory, T cells gamma delta, NK cells resting, NK cells activated, Monocytes, Macrophages M0, Macrophages M1, Macrophages M2, Dendritic cells resting, Dendritic cells activated, Mast cells resting, Mast cells activated, Eosinophils, Neutrophils. Multiple categories show asterisks above the boxes. Higher psoriasis fractions are visually prominent for Dendritic cells activated, Macrophages M0, Macrophages M1 and Neutrophils, with several points reaching near 0.6. Higher control fractions are visually prominent for Dendritic cells resting and Mast cells resting. B) A lollipop correlation plot titled CDC20. X-axis label: Correlation Coefficient, unit not shown, range negative 0.6 to positive 0.6. Y-axis lists immune cell types. Dot size encodes abs(cor) with labeled sizes 0.1, 0.2, 0.3, 0.4, 0.5. A pvalue legend ranges 0 to 1. Strongest positive correlations appear for Dendritic cells activated, T cells CD4 memory activated and T cells follicular helper, each near positive 0.6 with pvalue less than 0.001. Strongest negative correlations appear for Mast cells resting, T cells regulatory and Dendritic cells resting, each near negative 0.5 to negative 0.6 with pvalue less than 0.001. Additional pvalues printed include 0.004, 0.007, 0.013, 0.014, 0.043, 0.080, 0.110, 0.200, 0.212, 0.225, 0.311, 0.351, 0.374, 0.397, 0.783. The two graphs relate by showing group differences in immune cell fractions and the CDC20 correlation with those immune cell types.

Open in a new tab

Immune infiltration and key gene correlations. (A) Comparison of immune cell infiltration between control and psoriasis groups. **P < 0.01; ***P < 0.001. (B) Correlation between CDC20 and immune cell infiltration.

Construction of Core Gene Transcriptional Regulatory Network and Screening of Potential Targeted Drugs

Based on transcription factor (TF)–target gene regulatory relationships, we constructed a transcriptional regulatory network comprising eight core psoriasis-associated genes and 35 upstream TFs (Figure 10A). Network topology analysis revealed that CDK1 is targeted by multiple TFs, including ARID3A, ETS2, and TP53, playing a central role in cell cycle regulation; CCNA2 is regulated by SP1, MYC, and TP53, and is critical for cell cycle progression and proliferation; CCNB1 interacts with inflammation-related TFs such as FOXM1, NFKB1, and IRF1, contributing to both cell cycle control and immune-inflammatory responses. BIRC5, regulated by FOXO3 and HDAC1, is mainly involved in apoptosis resistance, while TOP2A and CDC20 are controlled by YBX1 and participate in cell cycle regulation. ASPM and CEP55 engage in cell division and proliferation through their respective transcriptional regulatory axes.

Figure 10.

A network diagram and a bubble scatter plot showing transcription factor regulation and drug enrichment ranking. Image A illustrates a transcription factor core gene regulatory network with lines connecting circular core genes and triangular transcription factors, indicating regulatory links. Core genes include BIRC5, TOP2A, CDC20, ASPM, CDK1, CCNB1, CCNA2 and CEP55. Transcription factors like TP53, MYC, SP1, FOXM1, NFKB1, IRF1, ARID3A, ETS2, FOXO3 and HDAC1 connect to multiple core genes, forming a dense network around CDK1, CCNB1, CCNA2 and BIRC5. Image B features a bubble scatter plot ranking drugs by key gene enrichment. The X-axis shows Gene ratio (0.3 to 1.0) and the Y-axis lists drugs, with LUCANTHONE at the top and mechlorethamine at the bottom. A cluster of points is around a gene ratio of 0.35 to 0.60, with LUCANTHONE near 1.0. The legend indicates p-values from 5.0e-07 to 2.0e-06 and bubble sizes of 3, 5, 7, 9 and 11. Both parts relate by showing a regulatory network and drug enrichment ranking linked to these genes.

Open in a new tab

Transcriptional regulation and drug enrichment analysis. (A) Transcription factor–core gene regulatory network. (B) Ranking of potential drugs based on key gene enrichment.

Subsequently, small-molecule drug enrichment analysis based on these core genes identified 30 potential targeted compounds (Figure 10B). These compounds predominantly modulate cell cycle progression, epigenetic modification, immune regulation, or exhibit anti-tumour activity. Notably, Lucanthone displayed the highest enrichment, with a gene ratio approaching 1.0 and an extremely low p-value, suggesting a strong and specific targeting potential against the core psoriasis genes.

Colocalization Analysis of Key Gene eQTLs with Psoriasis GWAS Signals

To further elucidate the genetic regulatory mechanisms of core psoriasis genes, we performed colocalization analysis of 11 key genes with expression quantitative trait loci (eQTLs) and psoriasis GWAS signals (Figure 11A). All genes exhibited significant eQTL–GWAS colocalization. Notably, CDK1 corresponded to rs72637065 on chromosome 10 at 62.4 Mb, BIRC5 to rs11655396 on chromosome 17 at 76.2 Mb, TOP2A to rs189383841 on chromosome 17 at 38.5 Mb, CCNB1 to rs73121291 on chromosome 5 at 68.4 Mb, NUSAP1 to rs140021954 on chromosome 15 at 41.7 Mb, KIF11 to rs4933735 on chromosome 10 at 94.4 Mb, CEP55 to rs41308663 on chromosome 10 at 95.2 Mb, CCNA2 to rs116167097 on chromosome 4 at 122.7 Mb, MELK to rs10973045 on chromosome 9 at 36.6 Mb, CDC20 to rs11210834 on chromosome 1 at 43.8 Mb, and ASPM to rs144723735 on chromosome 1 at 197.2 Mb.

Figure 11.

Scatter, Manhattan and Venn plots analyze gene colocalization and expression overlap. Panel A consists of three scatter plots analyzing GWAS and eQTL colocalization. The first plot shows GWAS -log10(P) versus eQTL -log10(P), highlighting rs11210834 with a color scale for r squared from 0.2 to 0.8, indicating potential colocalization. The second plot lacks an x-axis label but displays GWAS -log10(P) with rs11210834 marked, suggesting a relationship between signals. The third plot has chr1 (Mb) on the x-axis and eQTL -log10(P) on the y-axis, again highlighting rs11210834, indicating its significance in genetic regulation. Panel B is a Manhattan plot with chromosomes 1 to 22 on the x-axis and -log10(P) on the y-axis, showing peaks on specific chromosomes, notably chromosome 16, suggesting strong genetic associations. Panel C presents a Venn diagram showing overlap between logFC greater than 0 and SMRb greater than 0, with 11 (3.4 percent) unique to logFC, 3 (0.9 percent) overlapping and 307 (95.6 percent) unique to SMRb. Panel D shows another Venn diagram for logFC less than 0 and SMRb less than 0, with 0 (0 percent) unique to logFC, 3 (0.9 percent) overlapping and 337 (99.1 percent) unique to SMRb. The panels collectively illustrate genetic colocalization, association peaks and expression overlap, emphasizing rs11210834′s role in genetic studies.

Open in a new tab

Genetic colocalization and Mendelian randomization analysis. (A) Colocalization of key gene eQTLs with psoriasis GWAS signals. (B) Genome-wide association Manhattan plot. (C) Overlap visualization of up-regulated differentially expressed genes and genes with positive effect sizes from SMR analysis. (D) Overlap visualization of down-regulated differentially expressed genes and genes with negative effect sizes from SMR.

A genome-wide Manhattan plot was constructed to systematically identify differentially expressed genes associated with psoriasis (Figure 11B). Differentially expressed genes were broadly distributed across chromosomes 1–22, with CFAP58, GSTO1, CFAP58-DT, GSTO2, ITPRIP, and CD22 showing the most significant expression differences. Immune-inflammatory genes such as CXCL5, TNF, NLRP6, GNLY, and GZMM, as well as multiple genes related to epidermal structure and proliferation, also exhibited pronounced differential signals.

To assess the concordance between gene expression changes and genetic regulatory effects, we intersected upregulated and downregulated genes with SMR results showing positive or negative effect sizes, respectively (Figure 11C and D). Three upregulated genes—LYN, IL1RN, and TRBC1—and three downregulated genes—WDR47, CES1, and LMOD1—demonstrated consistent directionality between expression and genetic effect.

Causal relationships of these six genes were further evaluated using SMR effect plots and locus plots (Figure 12A and B). Effect plots confirmed that cis-eQTL effect sizes aligned with GWAS effect directions for all six genes. Chromosomal localization indicated LYN at chr8: 56.0–58.0 Mb, IL1RN at chr2: 113.0–115.0 Mb, TRBC1 at chr7: 141.5–143.5 Mb, WDR47 at chr1: 108.5–110.5 Mb, CES1 at chr16: 55.0–57.0 Mb, and LMOD1 at chr1: 201.0–203.0 Mb.

Figure 12.

Two plots of ENSG00000254087 showing aligned eQTL and GWAS effects and a chromosome 8 locus peak. Image A displays a scatter plot titled ENSG00000254087. The x-axis shows eQTL effect sizes from -0.13 to 0.13, while the y-axis shows GWAS effect sizes from -0.388 to 0.388. Triangles represent top cis-eQTLs and circles represent cis-eQTLs. Two clusters are visible: one near x=-0.13 to -0.10, y=-0.30 to 0 and another near x=0.10 to 0.13, y=0.05 to 0.38. A diagonal dashed line runs from lower left to upper right. A vertical color bar labeled r squared ranges from 0.17 to 1.00. Image B features a two-track locus plot with gene annotations. The top track y-axis is labeled -log10 of P (GWAS or SMR), ranging from 0 to 4, with a dotted line at 3 and text pSMR=0.001. The bottom track y-axis is labeled -log10 of P (eQTL), ranging from 0 to 110. The x-axis is Chromosome 8 Mb, ranging from 56.0 to 58.0. The strongest eQTL peak is near 56.8 to 56.9 Mb. Gene labels include SBFP1, LOC105375843, TMEM68, TGS1, LYN, SNORD54, MOS, CHCHD7, PLAG1, RPS20, SOR1C6CP and SOR1C6S.

Open in a new tab

Key gene validation and integrated model display. (A and B) Core gene expression and validation in mouse model.

Discussion

In the present study, we applied an integrated analytical framework combining differential expression analysis, weighted gene co-expression network analysis, machine learning, immune infiltration profiling, transcriptional regulatory analysis, drug enrichment, colocalization, and summary data-based Mendelian randomization to characterize the molecular features of psoriasis. Rather than supporting a single-gene model, our findings converge on a broader cell cycle–immune inflammation signature, in which proliferative keratinocyte programmes and inflammatory immune components appear to be closely coupled. This interpretation is consistent with the established pathological features of psoriasis, including epidermal hyperplasia, abnormal differentiation, and sustained immune activation.

A total of 950 differentially expressed genes were identified, and integration with the psoriasis-associated MEbrown module yielded 767 candidate genes. Functional enrichment analysis showed that these genes were mainly involved in cell cycle regulation, cytokine activity, chemokine-related functions, and immune-inflammatory pathways. Protein–protein interaction analysis further highlighted hub genes such as CCNA2, CCNB1, CDC20, and CDK1, suggesting that aberrant activation of cell cycle-related machinery may be a prominent transcriptional feature of psoriatic lesions. These findings support the view that excessive keratinocyte proliferation in psoriasis is not an isolated event, but part of a broader pathogenic programme linked to inflammatory signalling and tissue remodelling.

Among the prioritised genes, the Lasso–XGBoost model identified 11 candidate markers with strong discriminatory performance across datasets, with CDC20 showing the highest AUC value. However, these genes may be better interpreted as components of a proliferative disease-associated expression pattern rather than as standalone disease-specific drivers.13–15 In this context, CDC20, CCNB1, and CDK1 likely reflect enhanced mitotic activity and dysregulated epidermal turnover, both of which are central to psoriatic lesion development.16,17 The experimental validation performed in the imiquimod-induced mouse model further supported this interpretation, as the upregulation of these proteins was accompanied by typical histopathological changes, including epidermal hyperplasia and inflammatory infiltration. Together, these results suggest that cell cycle-related genes may serve as useful candidates for biomarker prioritization, while their precise mechanistic roles require further investigation.

Immune infiltration analysis provided an additional layer of evidence for the inflammatory context of psoriasis. Compared with controls, psoriatic lesions showed increased infiltration of several pro-inflammatory immune cell populations, including activated CD4 memory T cells, follicular helper T cells, γδ T cells, M0/M1 macrophages, activated dendritic cells, and neutrophils,18,19 whereas resting dendritic cells and resting mast cells were relatively reduced.20,21 These findings are consistent with the recognised contribution of both innate and adaptive immune responses to psoriasis pathogenesis. Notably, the prioritised genes showed strong positive correlations with activated dendritic cells and strong negative correlations with resting mast cells, suggesting that the transcriptional programme represented by these genes may be linked to immune activation status within lesional skin. This pattern supports a model in which abnormal keratinocyte proliferation and immune-cell-mediated inflammation are biologically intertwined rather than independent processes.22,23

The transcription factor regulatory network further reinforced this interpretation. Several of the key genes were linked to transcription factors known to participate in cell proliferation, inflammatory signalling, and stress responses.24,25 From a systems perspective, this suggests that the observed transcriptional changes in psoriasis may arise from coordinated upstream regulatory programmes rather than isolated gene-level abnormalities. Such an interpretation is more consistent with the complex and relapsing nature of psoriasis, in which epidermal, immune, and genetic factors interact dynamically.22,26 Accordingly, the prioritised genes should be viewed as nodes within a broader regulatory network that reflects the disease state.

Drug enrichment analysis identified several compounds potentially connected to the core gene set, with Lucanthone emerging as the top-ranked candidate in silico. Given its known associations with DNA damage response, proliferation control, and inflammatory signalling in other biological contexts, Lucanthone may represent a hypothesis-generating candidate for future evaluation in psoriasis.27–29 However, this result should be interpreted cautiously. In the absence of direct pharmacological validation in psoriasis-related experimental systems, the current findings do not support therapeutic conclusions, but rather provide a starting point for future drug repositioning studies.

The genetic analyses added another dimension to the interpretation of the results. Colocalization analysis indicated that the prioritised genes were associated with psoriasis-relevant eQTL–GWAS signals, while SMR analysis identified several genes, including LYN and IL1RN, with consistent directionality between gene expression and genetically predicted effects. Importantly, these genes were not identical to the top-ranked transcriptomic markers highlighted by the machine learning model. This distinction suggests that genes with strong discriminatory performance in transcriptomic datasets are not necessarily those with the strongest genetically supported associations.30–32 Therefore, the present work highlights two complementary but non-equivalent layers of prioritization: one based on disease-state expression patterns and another based on genetic regulatory support. This distinction may be important for future studies aiming to separate biomarkers of disease activity from genes more directly implicated in inherited susceptibility or regulatory architecture.33–35

Compared with studies relying on a single analytical strategy, the present work used a multi-step integrative design to connect transcriptomic alterations with immune characteristics and genetic signals. This approach improves robustness at the level of candidate prioritization and provides a broader view of psoriasis-associated molecular features. At the same time, the findings should be interpreted within the constraints of the study design.

The analyses were based primarily on public datasets, which may retain residual inter-cohort heterogeneity despite batch-effect correction. Machine-learning models also carry intrinsic over-fitting risks even with cross-validation. The short-term imiquimod-induced murine model only enables histological and protein-level expression assessment and cannot recapitulate the chronic nature of human psoriasis; the observed upregulation of cell-cycle-related proteins including CDC20 represents merely correlative evidence and does not establish their direct functional contribution to disease pathogenesis. Similarly, drug-enrichment outputs remain purely in-silico exploratory predictions without supporting wet-lab experimental validation, and genetic analyses provide only supportive evidence for candidate prioritization rather than definitive mechanistic proof. Moreover, the absence of a large independent human validation cohort limits the immediate translational applicability of the identified candidates. In addition, samples from other inflammatory skin diseases were not included in this study; therefore, the disease specificity of these genes cannot be confirmed.

Conclusion

In summary, this study identifies a psoriasis-associated cell cycle–immune inflammation signature and several candidate genes. CDC20, CCNB1, and CDK1 may serve as candidate disease-activity biomarkers, whereas the SMR-supported LYN and IL1RN may reflect genetic susceptibility to psoriasis. CDC20 should be considered a candidate hub biomarker rather than a causal driver. These findings provide potential biomarkers and directions for further mechanistic and therapeutic studies in psoriasis.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Ethics Approval

The bioinformatics component of this study was based exclusively on publicly available, anonymized, and non-identifiable human data. According to Item 2 of Article 32 of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects, issued by the relevant Chinese authorities on February 18, 2023, this component was exempt from additional ethical review and informed consent requirements. All animal care and experimental procedures were performed in strict accordance with the Guidelines for the Care and Use of Laboratory Animals of China and the ARRIVE guidelines. The animal experimental protocol was reviewed and approved by the Medical Ethics Committee of the Affiliated Hospital of Qinghai University (Ethics Approval No. FYXC-2024-13). All efforts were made to minimize animal pain, distress, and suffering.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors declare that they have no competing interests.

References

  • 1.Armstrong AW, Blauvelt A, Callis Duffin K, et al. Psoriasis. Nat Rev Dis Primers. 2025;11(1):45. doi: 10.1038/s41572-025-00630-5 [DOI] [PubMed] [Google Scholar]
  • 2.Helm MF, Clebak KT, Critchfield L, et al. Psoriasis and papulosquamous disorders. Prim Care. 2025;52(3):567–17. doi: 10.1016/j.pop.2025.05.007 [DOI] [PubMed] [Google Scholar]
  • 3.Zou Y, Liao K, Traidl S, et al. Burden of inflammatory skin diseases based on GBD data: China vs USA. Acta Derm Venereol. 2025;105:adv43753. doi: 10.2340/actadv.v105.43753 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Mou Y, Ma Y, Yu X, et al. Mitochondria-targeted nanozyme system for psoriasis treatment. J Nanobiotechnology. 2026;24(1):222. doi: 10.1186/s12951-026-04068-z [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Li Z, Deng Y, Xiao F, et al. Identification of novel lipid metabolism-related biomarkers of aortic dissection by integrating single-cell RNA sequencing analysis and machine learning algorithms. Front Immunol. 2025;16:1681989. doi: 10.3389/fimmu.2025.1681989 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Zhou X, Zhou H, Luo X, et al. Discovery of biomarkers in the psoriasis through machine learning and dynamic immune infiltration in three types of skin lesions. Front Immunol. 2024;15:1388690. doi: 10.3389/fimmu.2024.1388690 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Deng J, Schieler C, Borghans JAM, et al. Finding gene regulatory networks in psoriasis: application of a tree-based machine learning approach. Front Immunol. 2022;13:921408. doi: 10.3389/fimmu.2022.921408 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Strange A, Capon F, Spencer CC, et al. A genome-wide association study identifies new psoriasis susceptibility loci and an interaction between HLA-C and ERAP1. Nat Genet. 2010;42(11):985–990. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Si S, Wang X, Zhan S. Integrated multi-omics strategies for identifying novel therapies in psoriasis. Bioinformatics. 2026;42(6). doi: 10.1093/bioinformatics/btag347 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Li L, Wang H, Gao Y, et al. Integrative profiling of gut microbiome, bacteriophagenome, and predicted metabolome in obese adults: novel insights into intervention targets. BMC Microbiol. 2026;26(1):226. doi: 10.1186/s12866-025-04682-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Liu F, Wang S, Liu B, et al. (R)-Salbutamol improves imiquimod-induced psoriasis-like skin dermatitis by regulating the Th17/Tregs balance and glycerophospholipid metabolism. Cells. 2020;9(2):511. doi: 10.3390/cells9020511 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Okubo A, Uchida Y, Higashi Y, et al. CD147 is essential for the development of psoriasis via the induction of Th17 cell differentiation. Int J Mol Sci. 2021;23(1):177. doi: 10.3390/ijms23010177 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Chang X, Tamauchi S, Nakagawa A, et al. Chemotherapeutic hormesis induced by the tumor microenvironment in refractory ovarian cancer. Sci Rep. 2025;15(1):596. doi: 10.1038/s41598-024-84290-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Mukherjee A, Bhattacharyya J, Sagar MV, et al. Liposomally encapsulated CDC20 siRNA inhibits both solid melanoma tumor growth and spontaneous growth of intravenously injected melanoma cells on mouse lung. Drug Deliv Transl Res. 2013;3(3):224–234. doi: 10.1007/s13346-013-0141-3 [DOI] [PubMed] [Google Scholar]
  • 15.Yamashita N, Yoshizuka A, Kase A, et al. Activation of the aryl hydrocarbon receptor by 3-methylcholanthrene, but not by indirubin, suppresses mammosphere formation via downregulation of CDC20 expression in breast cancer cells. Biochem Biophys Res Commun. 2021;570:131–136. doi: 10.1016/j.bbrc.2021.07.047 [DOI] [PubMed] [Google Scholar]
  • 16.Melero JL, Andrades S, Arola L, et al. Deciphering psoriasis. A bioinformatic approach. J Dermatol Sci. 2018;89(2):120–126. doi: 10.1016/j.jdermsci.2017.11.010 [DOI] [PubMed] [Google Scholar]
  • 17.Wu Z, Wang R, Liu Y, et al. RNA sequencing reveals that the genes related to cell cycle and glycolysis play an essential role in IL-27-induced keratinocyte hyperproliferation. J Inflamm Res. 2024;17:8165–8180. doi: 10.2147/JIR.S481835 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Su W, Wei Y, Huang B, et al. Identification of hub genes and immune infiltration in psoriasis by bioinformatics method. Front Genet. 2021;12:606065. doi: 10.3389/fgene.2021.606065 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Zhang Y, Shi Y, Lin J, et al. Immune cell infiltration analysis demonstrates excessive mast cell activation in psoriasis. Front Immunol. 2021;12:773280. doi: 10.3389/fimmu.2021.773280 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Dong Y, Ge HS, Chang RX, et al. Potential marker genes for psoriasis revealed based on single-cell sequencing and mendelian randomization analysis. Front Genet. 2025;16:1634874. doi: 10.3389/fgene.2025.1634874 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Li S, Qi Y, Qiao D, et al. Machine learning meets psoriasis: identifying key lactylation biomarkers as potential targets for diagnosis and therapies. Front Immunol. 2026;17:1791693. doi: 10.3389/fimmu.2026.1791693 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Kamata M, Tada Y. Crosstalk: keratinocytes and immune cells in psoriasis. Front Immunol. 2023;14:1286344. doi: 10.3389/fimmu.2023.1286344 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Tonel G, Conrad C. Interplay between keratinocytes and immune cells--recent insights into psoriasis pathogenesis. Int J Biochem Cell Biol. 2009;41(5):963–968. doi: 10.1016/j.biocel.2008.10.022 [DOI] [PubMed] [Google Scholar]
  • 24.Goldminz AM, Au SC, Kim N, et al. NF-κB: an essential transcription factor in psoriasis. J Dermatol Sci. 2013;69(2):89–94. doi: 10.1016/j.jdermsci.2012.11.002 [DOI] [PubMed] [Google Scholar]
  • 25.Sano S, Chan KS, Carbajal S, et al. Stat3 links activated keratinocytes and immunocytes required for development of psoriasis in a novel transgenic mouse model. Nat Med. 2005;11(1):43–49. doi: 10.1038/nm1162 [DOI] [PubMed] [Google Scholar]
  • 26.Rendon A, Schäkel K. Psoriasis Pathogenesis and Treatment. Int J Mol Sci. 2019;20(6):1475. doi: 10.3390/ijms20061475 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Carew JS, Espitia CM, Esquivel JA, et al. Lucanthone is a novel inhibitor of autophagy that induces cathepsin D-mediated apoptosis. J Biol Chem. 2011;286(8):6602–6613. doi: 10.1074/jbc.M110.151324 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Luo M, Kelley MR. Inhibition of the human apurinic/apyrimidinic endonuclease (APE1) repair activity and sensitization of breast cancer cells to DNA alkylating agents with lucanthone. Anticancer Res. 2004;24(4):2127–2134. [PubMed] [Google Scholar]
  • 29.Radin DP, Smith G, Moushiaveshi V, et al. Lucanthone targets lysosomes to perturb glioma proliferation, chemoresistance and stemness, and slows tumor growth in vivo. Front Oncol. 2022;12:852940. doi: 10.3389/fonc.2022.852940 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Giambartolomei C, Vukcevic D, Schadt EE, et al. Bayesian test for colocalisation between pairs of genetic association studies using summary statistics. PLoS Genet. 2014;10(5):e1004383. doi: 10.1371/journal.pgen.1004383 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Hormozdiari F, van de Bunt M, Segrè AV, et al. Colocalization of GWAS and eQTL signals detects target genes. Am J Hum Genet. 2016;99(6):1245–1260. doi: 10.1016/j.ajhg.2016.10.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Zhu Z, Zhang F, Hu H, et al. Integration of summary data from GWAS and eQTL studies predicts complex trait gene targets. Nat Genet. 2016;48(5):481–487. doi: 10.1038/ng.3538 [DOI] [PubMed] [Google Scholar]
  • 33.Dand N, Stuart PE, Bowes J, et al. GWAS meta-analysis of psoriasis identifies new susceptibility alleles impacting disease mechanisms and therapeutic targets. Nat Commun. 2025;16(1):2051. doi: 10.1038/s41467-025-56719-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Jeong Y, Song J, Lee Y, et al. A transcriptome-wide analysis of psoriasis: identifying the potential causal genes and drug candidates. Int J Mol Sci. 2023;24(14):11717. doi: 10.3390/ijms241411717 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Zhu D, Yao S, Wu H, et al. A transcriptome-wide association study identifies novel susceptibility genes for psoriasis. Hum Mol Genet. 2021;31(2):300–308. doi: 10.1093/hmg/ddab237 [DOI] [PubMed] [Google Scholar]

Articles from Clinical, Cosmetic and Investigational Dermatology are provided here courtesy of Dove Press

RESOURCES