Abstract: BACKGROUND: Rheumatoid arthritis and Crohn’s disease are common autoimmune diseases. Clinical studies have found that these two diseases can coexist and may be related, but there is currently no research to prove that there are common pathogenic genes and immune mechanisms between them.
OBJECTIVE: To identify the shared genes and immune mechanisms between rheumatoid arthritis and Crohn’s disease through bioinformatics and two machine learning methods.
METHODS: Training and validation datasets for rheumatoid arthritis and Crohn’s disease were retrieved from the GEO database (an open database developed by the United States National Library of Medicine) and uniformly organized. The “limma” package was used to perform differentially expressed genes of rheumatoid arthritis and Crohn’s disease. Weighted gene co-expression network analysis was applied to the training sets of rheumatoid arthritis and Crohn’s disease to identify disease-related modules, and the intersection was taken to preliminarily screen gene sets, while GO and KEGG analyses were conducted. Twenty gene sets were identified through the protein-protein interaction network and MCODE algorithm. Two machine learning algorithms, LASSO regression and random forest, were independently applied to the training sets of rheumatoid arthritis and Crohn’s disease to screen key characteristic genes for each disease. Subsequently, the intersection of the screening results of rheumatoid arthritis and Crohn’s disease was taken to obtain shared potential key genes, and the accuracy was verified through the validation set to determine the core genes. Finally, CIBERSORT immune infiltration and other functional analyses were performed to confirm the correlation between core genes and rheumatoid arthritis as well as Crohn’s disease.
RESULTS AND CONCLUSION: A total of 2 516 differentially expressed genes were obtained for rheumatoid arthritis, and 281 differentially expressed genes for Crohn’s disease. Following intersection analysis using WGCNA, protein-protein interaction network, and two machine learning algorithms, three core genes were identified: CASP1, TRIM21, and PSMB10. Enrichment analysis showed that the two diseases were associated with antigen processing and presentation, luminal side of endoplasmic reticulum membrane, and binding to multiple immunoglobulins. The expression trends of three core genes in the validation sets of the two diseases were consistent with those in the training set. Immune cell infiltration analysis revealed significantly increased expression of M0 macrophages, M1 macrophages, and neutrophils in both rheumatoid arthritis and Crohn’s disease. This indicates that neutrophils may play an important role in the pathogenesis of rheumatoid arthritis and Crohn’s disease. This study not only enhances our understanding of the commonalities between two important autoimmune diseases, but more importantly, it provides valuable insights and specific guidance for Chinese researchers in the discovery and validation of new targets, the development of novel diagnostic and therapeutic technologies, and the application of translational medicine research models.

Key words: rheumatoid arthritis, Crohn’s disease, immune infiltration, bioinformatics, machine learning, autoimmune disease