Identification of potential candidate genes in rheumatoid arthritis using integrated machine learning and WGCNA approach on transcriptomic data

Rheumatoid Arthritis is a multifactorial systemic autoimmune disease and a significant cause of morbidity, mortality and poor life quality. Gene expression data from the GEO database is used in our study. We first identified relevant genes from significant modules using weighted gene co-expression network analysis (WGCNA), created a PPI network and then employed machine learning algorithms to find feature genes. A wide range of Bioinformatic tools were utilized ranging from clusterProfiler for functional enrichment analysis, gene set enrichment analysis tool to identify biological important functions, cibersort for immune infiltration analysis and DEGGs to identify differentially expressed gene–gene interactions. Finally, FDA-approved anti-rheumatic drugs were docked against selected target regions. Our findings unveil two potential intersecting biomarkers IFIT3 and IFIT2 via MLSeq and WGCNA analysis. They were shown to be closely related to the high concentration of specific immune cell type such as neutrophils in the patient group. The GSEA analysis showed that the oxidative phosphorylation pathway was significantly enriched in downregulation. Finally, molecular docking results showed Anakinra and Methotrexate as the best candidate drugs that might suppress the expression of potential candidate RA-associated proteins. Our study concluded by identifying potential biomarkers for RA that could be considered for clinical validation. These biomarkers would also provide a solid basis for a thorough investigation of potential RA-associated pathways and the discovery of new therapeutic targets that could significantly influence the disease's onset and progression.