Integrated analysis of single-cell and bulk RNA-sequence data reveal plasma dendritic cell-related diagnostic model for Sjögren's syndrome based on machine learning.

Background/purpose: Plasmacytoid dendritic cells (pDCs) play a critical role in linking innate and adaptive immunity in the pathogenesis of Sjögren's syndrome (SS). This study aimed to characterize pDC functional features in SS by integrating single-cell and bulk transcriptomic data and to develop a pDC-related gene signature-based diagnostic model. Materials and methods: Single-cell RNA sequencing data were analyzed using quality control, clustering, and single-cell weighted gene co-expression network analysis (scWGCNA) to identify pDC-associated gene modules. Bulk transcriptomic datasets were then used to screen diagnostic biomarkers through 204 combinations of 15 machine-learning algorithms, and a nomogram model was constructed. Cell-cell communication and drug-gene interaction analyses were subsequently performed. Results: A gene module strongly correlated with pDCs was identified, from which ten core biomarkers were selected to establish the diagnostic model. The model demonstrated excellent diagnostic performance, with AUCs of 0.999 in the training set and 0.866 in an independent validation set. Cell-cell communication analysis revealed active macrophage migration inhibitory factor (MIF) signaling between pDCs and monocyte subsets. Drug network and molecular docking analyses suggested that methotrexate and hydroxychloroquine may interact with proteins encoded by several biomarker genes. Conclusion: This study developed a robust SS diagnostic model based on pDC-related gene signatures and revealed potential pDC-monocyte interactions. The identified biomarkers and candidate drugs may facilitate auxiliary diagnosis and targeted therapeutic development for SS.