Combining WGCNA and machine learning to identify mechanisms and biomarkers of hyperthyroidism and atrial fibrillation

Linyuan WangKun YangRuilong KangPengbo LiuYongzhi Deng^*

• Department of Cardiovascular Surgery, The Affiliated Hospital of Shanxi Medical University, Shanxi Cardiovascular Hospital (Institute), Taiyuan, China

Background: Hyperthyroidism and atrial fibrillation (AF) are interrelated conditions with significant cardiovascular impact. While their clinical association is established, the molecular mechanisms remain unclear. Identifying shared biomarkers and pathways can advance understanding and guide therapy. Methods: The hyperthyroidism dataset GSE71956 and the AF dataset GSE115574 were obtained from the Gene Expression Omnibus (GEO) database. Differential gene analysis was performed using the "limma" package, and overlapping genes shared by both diseases were identified through weighted gene co-expression network analysis (WGCNA), followed by functional enrichment analysis. Machine learning algorithms were also applied to identify key biomarkers. To validate the predictive results, peripheral blood samples were collected for real-time quantitative polymerase chain reaction (RT-qPCR) analysis. Finally, immune infiltration analysis was conducted to evaluate immune cell changes in hyperthyroidism and AF. Results: Through differential gene screening and WGCNA, 23 overlapping genes associated with hyperthyroidism and AF were identified. Using least absolute shrinkage and selection operator (LASSO) and random forest (RF) machine learning algorithms, CXCL16 and TMEM127 were ultimately identified as key genes. The two genes demonstrated good diagnostic efficacy in the hyperthyroidism validation set GSE276271 (AUC: TMEM127, 0.636; CXCL16, 0.591) and in the AF validation set GSE2240 (AUC: TMEM127, 0.745; CXCL16, 0.720). RT – qPCR analysis demonstrated that CXCL16 and TMEM127 expression levels were significantly elevated in both the hyperthyroidism and AF groups compared to the control group, aligning with the findings from our prior bioinformatics analysis. Immune analysis revealed significant differences in two immune cell types in both hyperthyroidism and AF. Conclusion: CXCL16 and TMEM127 are promising biomarkers, offering insights into the shared pathogenesis of hyperthyroidism and AF. These findings provide a foundation for novel diagnostic and therapeutic strategies targeting these conditions.