HTR2A DNA Methylation as a Diagnostic Biomarker for Rheumatoid Arthritis: A Validation Study Using Targeted Sequencing and Machine Learning Algorithms

Abstract

Objectives To validate the potential of HTR2A cg15692052 DNA methylation as a diagnostic biomarker for RA and its subtypes. Methods MethylTarget™ targeted region methylation sequencing technology was employed to analyze the DNA methylation levels of HTR2A cg15692052 in RA, HC, ankylosing spondylitis (AS), psoriatic arthritis (PSA), gout, systemic lupus erythematosus (SLE), dermatomyositis (DM), and primary Sjögren's syndrome (SS) patients within the region of chr13:46898190 ~ chr13:46897976, spanning a total of 215bp. Logistic regression, LASSO, random forests, and Xgboost algorithms were used in R software to screen for significant variables, construct models, visualize results, and perform statistical analysis. Multiple imputation was applied to handle missing values, and Spearman's method was used to calculate correlations. Results Compared to the HC group, RA patients and four serological subtypes of RA (RF-negative RA, RF/CCP double-positive, RF/CCP double-negative, and CCP-negative RA) exhibited significantly higher levels of HTR2A cg15692052 methylation at positions 75/125/143/149/163/185/187 and in average methylation (P < 0.05). Methylation levels at all positions and average methylation in RA patients and its four serological subtypes were significantly positively correlated with erythrocyte sedimentation rate (ESR) or C-reactive protein (CRP) (P < 0.05). HTR2A cg15692052 displayed various haplotypes with differential proportions, among which the CCCCCCC haplotype was significantly elevated in RA (P < 0.05) and positively correlated with ESR and CRP (r = 0.13 and 0.21, P = 0.001 and P < 0.001). Conversely, the TTTTTTT haplotype was significantly decreased in RA (P < 0.05) and negatively correlated with CRP (r=-0.15, P = 0.002). Predictive models constructed using different machine learning algorithms, incorporating methylation levels of HTR2A cg15692052 at various positions combined with different clinical features, were able to significantly distinguish RA patients with AUCs ranging from 0.672 to 0.757, RF/CCP double-negative patients with AUCs from 0.825 to 0.966, RF/CCP double-positive RA patients with AUCs from 0.714 to 0.846, and RF-negative RA patients with AUCs from 0.928 to 0.932. Conclusions The DNA methylation level of HTR2A cg15692052 is associated with RA and can serve as a diagnostic biomarker for RA and its subtypes.