Construction and validation of a diagnostic model for dermatomyositis based on the LASSO algorithm.

Abstract

Introduction: Dermatomyositis (DM) is the most prevalent disease among myositis patients. The immune response is crucial in DM development. Bioinformatics research on immune-related genes in DM is limited. This study attempted to construct a diagnostic model and investigate immune characteristics of immune-related differentially expressed genes (DEGs), which could aid in DM diagnosis.

Material and methods: GSE46239 and GSE39454 datasets were from the GEO database, and batch effects were eliminated for use as the DM training set. DEG were identified and enrichment analysis was conducted between DM and normal samples. Intersection of DEGs and immune-related genes yielded immune-related DEGs, which were utilized to generate a PPI network. The diagnostic model was built by the LASSO method. The diagnostic model and effectiveness of model genes were evaluated through GSE143323. The correlation between immune cell infiltration in DM and diagnostic genes was analyzed. Finally, expression levels of HLA genes in DM and their correlation with diagnostic genes were examined.

Results: A total of 350 DEGs were identified. Seventy-one immune-related DEGs were screened. LASSO regression identified 5 immune-related DEGs (ACKR1, DHX58, IRF7, ISG15, and PSMB8) for constructing the DM diagnostic model. The model showed good effectiveness in training and validation sets (AUC of 0.99 and 0.958, respectively), and 5 immune-related DEGs also exhibited good effectiveness (AUC > 0.784). Diagnostic genes in DM were associated with M1 macrophages, M2 macrophages, resting dendritic cells, and certain HLA genes.

Conclusions: We constructed a DM diagnostic model using ACKR1, DHX58, IRF7, ISG15, and PSMB8, which were closely related to immune cells and HLA. This model could contribute to research in DM diagnosis.