Data and code for: Integration of single cell and bulk transcriptomics identifies a myeloid specific five gene diagnostic signature and ceRNA network for psoriatic arthritis

Abstract BackgroundPsoriatic arthritis (PsA) is a chronic inflammatory disease affecting approximately 25–30% of patients with psoriasis, yet diagnostic delays remain common due to the lack of specific biomarkers. Myeloid cells play a central role in PsA pathogenesis, but a systematic myeloid‑centric characterization at the single‑cell level has not been fully established. This study aimed to identify a myeloid‑specific diagnostic gene signature and explore its regulatory mechanisms through integrated transcriptomic analysis. MethodsWe integrated single‑cell RNA‑seq data from PsA synovial fluid (GSE277596) with four bulk transcriptomic datasets from the Gene Expression Omnibus. Differentially expressed genes between PsA and controls were identified in myeloid subpopulations and in the training set GSE61281. A logistic regression model was constructed and evaluated using receiver operating characteristic curves, decision curve analysis, and calibration plots. A competing endogenous RNA (ceRNA) network and a four‑layer regulatory network (transcription factor–circRNA–miRNA–mRNA) were constructed using CircInteractome, miRNet, and dorothea. Immune infiltration was estimated with xCell, pseudotime trajectory was inferred with slingshot, and Mendelian randomisation was performed using eQTLGen and PsA genome‑wide association study data. ResultsA five‑gene signature (PSMA4, HINT1, LYN, RPS15A, HNMT) was identified, with genes predominantly expressed in proliferating myeloid cells and inflammatory macrophages. The logistic regression model showed good discriminative ability in the training set (area under the curve [AUC] = 0.942) and in three independent validation cohorts (AUC ≥ 0.8665). A ceRNA network comprising 20 circRNAs, 52 miRNAs and the five mRNAs was constructed, and the hsa_circ_0006022 / miR‑520f‑3p / PSMA4 axis was highlighted as a candidate regulatory module. Mendelian randomisation suggested a causal protective effect of PSMA4 expression on PsA risk (odds ratio = 0.26, P = 1.8 × 10⁻¹⁰). Immune infiltration analysis revealed enrichment of CD4⁺ memory T cells, and pseudotime analysis showed dynamic expression patterns along myeloid differentiation. ConclusionsThis study established a myeloid‑specific ceRNA network and a five‑gene diagnostic model for PsA through integrative transcriptomic analysis. The findings provide a hypothesis‑generating signature for future diagnostic development, with the hsa_circ_0006022 / miR‑520f‑3p / PSMA4 axis emerging as a candidate for mechanistic exploration. All results are computational and require experimental and prospective clinical validation.