# Unveiling potential diagnostic biomarkers for rheumatoid arthritis through integrated gene expression analysis

## TL;DR

This study identifies key genes and potential drugs for rheumatoid arthritis by analyzing gene expression data and validating results in patient tissues.

## Contribution

The study proposes novel diagnostic biomarkers and therapeutic targets for rheumatoid arthritis using integrated gene expression and machine learning analysis.

## Key findings

- 273 key genes were identified through WGCNA and DEG analysis, primarily involved in inflammatory pathways.
- Five core genes (GABARAPL1, FKBP5, PCDH9, and SLAMF8) were validated as potential diagnostic biomarkers for RA.
- Three drugs—(+)-chelidonine, daunorubicin, and bisacodyl—were predicted to target these key genes.

## Abstract

Rheumatoid arthritis (RA) is a chronic autoimmune disorder that significantly impacts quality of life. Despite extensive research, its pathogenesis remains unclear. This study aims to identify potential diagnostic biomarkers and therapeutic targets for RA.

This study integrated patient data from three Gene Expression Omnibus (GEO) databases to analyze gene expression in RA. Using Weighted Gene Correlation Network Analysis (WGCNA), we identified key genes, which were then compared with differentially expressed genes (DEGs) to uncover RA-related genes. Functional enrichment analysis provided insights into the biological roles of these genes. To refine our findings, we applied three algorithms—RandomForest, SVM-REF, LASSO, and Convolutional Neural Networks (CNN)—to pinpoint a subset of core genes. We evaluated their diagnostic potential through receiver operating characteristic (ROC) curves and selected the top five genes with the highest area under the curve (AUC) values for constructing a predictive nomogram model. An interaction analysis was performed to investigate the relationship between these genes and immune cell infiltration. Finally, the expression of these core genes was validated in the synovial tissues of RA patients. Drug-protein interaction relationships were predicted using the DSigDB database.

Differential expression analysis identified 543 DEGs. We subsequently applied WGCNA to compare these DEGs with significant module genes, resulting in the identification of 273 key genes. Functional enrichment analysis indicated that these genes were primarily involved in inflammatory response pathways. Further analysis using four machine learning algorithms identified 11 core genes. Of these, the five genes with the highest AUC values were selected to construct a robust nomogram model. Immune infiltration analysis revealed significant differences in immune cell levels and pathways between RA patients and healthy controls, which were correlated with the expression of these five genes. Validation through quantitative real-time PCR (qRT-PCR), Western blot, and immunofluorescence (IF) confirmed that GABARAPL1, FKBP5, and PCDH9 expression was lower in RA synovial tissues compared to healthy controls, while SLAMF8 expression was elevated. Additionally, potential therapeutic drugs targeting these key genes, including (+)-chelidonine, daunorubicin, and bisacodyl, were predicted.

GABARAPL1, FKBP5, PCDH9, and SLAMF8 are identified as potential biomarkers for RA, offering insights into future therapeutic strategies.