Novel peripheral blood diagnostic biomarkers screened by machine learning algorithms in ankylosing spondylitis

Abstract

Background:Ankylosing spondylitis (AS) is a chronic inflammatory disorder of unknown etiology that is hard to diagnose early. Therefore, it is imperative to explore novel biomarkers that may contribute to the easy and early diagnosis of AS.Methods:Common differentially expressed genes between normal people and AS patients in GSE73754 and GSE25101 were screened by machine learning algorithms. A diagnostic model was established by the hub genes that were screened. Then, the model was validated in several data sets.Results:IL2RBandZDHHC18were screened using machine learning algorithms and established as a diagnostic model. Nomograms suggested that the higher the expression ofZDHHC18, the higher was the risk of AS, while the reverse was true forIL2RB in vivo. C-indexes of the model were no less than 0.84 in the validation sets. Calibration analyses suggested high prediction accuracy of the model in training and validation cohorts. The area under the curve (AUC) values of the model in GSE73754, GSE25101, GSE18781, and GSE11886 were 0.86, 0.84, 0.85, and 0.89, respectively. The decision curve analyses suggested a high net benefit offered by the model. Functional analyses of the differentially expressed genes indicated that they were mainly clustered in immune response–related processes. Immune microenvironment analyses revealed that the neutrophils were expanded and activated in AS while some T cells were decreased.Conclusion:IL2RBandZDHHC18are potential blood biomarkers of AS, which might be used for the early diagnosis of AS and serve as a supplement to the existing diagnostic methods. Our study deepens the insight into the pathogenesis of AS.