Memory CD4+T cell profile is associated with unfavorable prognosis in IgG4-related disease: Risk stratification by machine-learning

Yuxue Nie,Zheng Liu,Wei Cao,Yu Peng,Hui Lu,Ruijie Sun,Jingna Li,Linyi Peng,Jiaxin Zhou,Yunyun Fei,Mengtao Li,Xiaofeng Zeng,Wen Zhang,Taisheng Li
DOI: https://doi.org/10.1016/j.clim.2023.109301
IF: 10.19
2023-03-01
Clinical Immunology
Abstract:IgG4-related disease (IgG4-RD) is a chronic immune-mediated disease with heterogeneity. In this study, we used machine-learning approaches to characterize the immune cell profiles and to identify the heterogeneity of IgG4-RD. The XGBoost model discriminated IgG4-RD from HCs with an area under the receiver operating characteristic curve of 0.963 in the testing set. There were two clusters of IgG4-RD by k-means clustering of immunological profiles. Cluster 1 featured higher proportions of memory CD4<sup>+</sup>T cell and were at higher risk of unfavorable prognosis in the follow-up, while cluster 2 featured higher proportions of naïve CD4<sup>+</sup>T cell. In the multivariate logistic regression, cluster 2 was shown to be a protective factor (OR 0.30, 95% CI 0.10-0.91, P = 0.011). Therefore, peripheral immunophenotyping might potentially stratify patients with IgG4-RD and predict those patients with a higher risk of relapse at early time.
immunology
What problem does this paper attempt to address?