Comprehensive molecular analyses of an autoimmune-related gene predictive model and immune infiltrations using machine learning methods in moyamoya disease

Abstract

Background: Growing evidence suggests the links between moyamoya disease (MMD) and autoimmune diseases. However, the molecular mechanism from genetic perspective remains unclear. This study aims to clarify the potential roles of autoimmune-related genes (ARGs) in the pathogenesis of MMD.Methods: Two transcription profiles (GSE157628 and GSE141025) of MMD were downloaded from GEO databases. ARGs were obtained from the Gene and Autoimmune Disease Association Database (GAAD) and DisGeNET databases. Differentially expressed ARGs (DEARGs) were identified using “limma” R packages. GO, KEGG, GSVA, and GSEA analyses were conducted to elucidate the underlying molecular function. There machine learning methods (LASSO logistic regression, random forest (RF), support vector machine-recursive feature elimination (SVM-RFE)) were used to screen out important genes. An artificial neural network was applied to construct an autoimmune-related signature predictive model of MMD. The immune characteristics, including immune cell infiltration, immune responses, and HLA gene expression in MMD, were explored using ssGSEA. The miRNA-gene regulatory network and the potential therapeutic drugs for hub genes were predicted.Results: A total of 260 DEARGs were identified in GSE157628 dataset. These genes were involved in immune-related pathways, infectious diseases, and autoimmune diseases. We identified six diagnostic genes by overlapping the three machine learning algorithms: CD38, PTPN11, NOTCH1, TLR7, KAT2B, and ISG15. A predictive neural network model was constructed based on the six genes and presented with great diagnostic ability with area under the curve (AUC) = 1 in the GSE157628 dataset and further validated by GSE141025 dataset. Immune infiltration analysis showed that the abundance of eosinophils, natural killer T (NKT) cells, Th2 cells were significant different between MMD and controls. The expression levels of HLA-A, HLA-B, HLA-C, HLA-DMA, HLA-DRB6, HLA-F, and HLA-G were significantly upregulated in MMD. Four miRNAs (mir-26a-5p, mir-1343-3p, mir-129-2-3p, and mir-124-3p) were identified because of their interaction at least with four hub DEARGs.Conclusion: Machine learning was used to develop a reliable predictive model for the diagnosis of MMD based on ARGs. The uncovered immune infiltration and gene-miRNA and gene-drugs regulatory network may provide new insight into the pathogenesis and treatment of MMD.

Mass cytometry revealed the circulating immune cell landscape across different Suzuki stages of Moyamoya disease

“…This finding suggests a notable immune response specific to MMD, highlighting the potential role of immune mechanisms in the pathology of the disease [10]. Additionally, a close association was observed between dysregulated genes in MMD and genes related to autoimmune diseases [11]. Concerning peripheral immunity in MMD, previous single-cell sequencing has suggested that the interaction of peripheral immune cells may play a role in the pathogenesis of MMD [12].…”

Recent Advances for Global Perspectives on Etiology, Pathophysiology, Clinical Presentations, and Management of Moyamoya Disease

“…This includes identifying differentially expressed genes (DEGs) and immune complexes. The study also found that gene-miRNA interactions play a role in modifying protein expression during disease progression by targeting specific factors [37].…”

Bioinformatics analysis reveals the landscape of immune cell infiltration and novel immune-related biomarkers in moyamoya disease

“…They have the potential to be diagnostic biomarkers for MMD. Among them, PTPN11 showed a low expression pattern and the highest diagnostic accuracy, and it also showed the highest degree according to PPI network analysis, which was identified by Li et al (2022) , as aforementioned. PTPN11 encodes a non-receptor tyrosine phosphatase with 2 Src-homology 2 domains (Shp2) as well, and it was identified as the first proto-oncogene.…”

“…However, there is still no well-recognized biomarker for accurately diagnosing and predicting the outcome of MMD. CD38, PTPN11, NOTCH1, TLR7, KAT2B, and ISG15 were identified using machine learning, and a reliable diagnostic model was developed by Li et al (2022 ). Except RNF213, UNC13D was previously identified as a key biomarker with good specificity and sensitivity ( Jin and Duan, 2022 ).…”

“…Although immune infiltration was not typical in the pathology of MMD arteries, Masuda et al did report immune cell infiltrating components such as macrophages or T cells in the thickened vascular wall ( Masuda et al, 1993 ). Recently, Li et al reported the significant difference of immune infiltration in eosinophils, natural killer T (NKT) cells, and Th2 cells between MMD and controls ( Li et al, 2022 ), whereas Jin and Duan (2022) demonstrated an elevated abundance of neutrophils, monocytes, and natural killer cells in MMD. However, the detailed mechanism of immune cell function in MMD was still largely unknown.…”