Machine learning-based characterization of cuprotosis-related biomarkers and immune infiltration in Ankylosing spondylitis

Abstract

Background: Ankylosing spondylitis (AS) is a chronic inflammatory disease that mainly affects the axial skeleton. Meanwhile, copper is a mineral nutrient involved in cell proliferation and death pathways and has been observed in various diseases. Methods: The purpose of this study was to identify potential new biomarkers of ankylosing spondylitis through biomarker analysis and explore its immune cell infiltration analysis. Gene expression profiles of GSE73754, GSE25101 and GSE18781 datasets were retrieved from the GEO database. Three machine learning algorithms, i.e., minimum absolute shrinkage and selection operator ( LASSO ), random forest ( RF ) and support vector machine recursive feature elimination ( SVM-RFE ), were used to screen cuprotosis-related genes ( CRGs ) of ankylosing spondylitis. Besides, ankylosing spondylitis specimens were classified using the consensus clustering method. The three machine learning results were intersected with the CRGs obtained by difference analysis and the differential CRGs obtained by clustering to obtain key genes. Additionally, the infiltration of immune cells was detected by ssGSEA, and the cuprotosis-related genes related to immune cells and immune function were detected by spearman correlation analysis. The nomogram model was established, and the accuracy of cuprotosis-related genes in predicting the development of AS disease was verified by external data sets. Result: The two AS libraries in the GEO database were merged after removing the batch effect. Through the difference analysis, 6 CRGs were obtained, 9 CRGs were obtained by LASSO algorithm, 4 CRGs were obtained by SVM-RFE, and 13 CRGs were obtained by RF. It was also found by clustering the consistency of AS samples that they could be divided into 2 subtypes, and 5 CRGs were obtained as well. Finally, two key genes of SLC31A1 and PDHB were obtained by VENN intersection. The immune infiltration and GSEA analysis showed that PDHB had the strongest positive correlation with T cells CD8, followed by T cells CD4 memory resting, and had the strongest negative correlation with Neutrophils. Besides, SLC31A1 had the strongest positive correlation with NK cells resting and the strongest negative correlation with T cells CD8. By establishing a nomogram model, it was found that the two genes could better predict the development of AS, and the verification of the external data set GSE18781 indicated the stronger predictive ability of PDHB. Conclusion: In summary, the present study revealed several newly discovered cuprotosis-related genes that interfered with the progression of ankylosing spondylitis through immune infiltration, and found the large research space of PDHB.