The Potential Diagnostic Value of Immune-Related Genes in Interstitial Fibrosis and Tubular Atrophy after Kidney Transplantation

Abstract

Background. Inflammation within areas of interstitial fibrosis and tubular atrophy (IF/TA) is associated with kidney allograft failure. The aim of this study was to reveal new diagnostic markers of IF/TA based on bioinformatics analysis. Methods. Raw data of IF/TA samples after kidney transplantation and control samples after kidney transplantation were extracted from the Gene Expression Omnibus (GEO) database (GSE76882 and GSE120495 datasets), and genes that were differentially expressed between the two groups (DEGs) were screened. Gene Set Enrichment Analysis (GSEA), ESTIMATE and single sample GSEA (ssGSEA), least absolute shrinkage and selection operator (LASSO) regression analysis, and competing endogenous RNA (ceRNA) network were used to analyze the data. Results. The results of GSEA revealed that multiple immune-related pathways were enriched in the IF/TA group, and subsequent immune landscape analysis also showed that the IF/TA group had higher immune and stromal scores and up to 15 types of immune cells occupied them, such as B cells, cytotoxic cells, and T cells. LASSO regression analysis selected 6 (including ANGPTL3, APOH, LTF, FCGR2B, HLA-DQA2, and EGF) out of 14 DE-IRGs as diagnostic genes to construct a diagnostic model. Then, receiver operating characteristic (ROC) curve analysis showed the powerful diagnostic value of the model, and the area under the curve (AUC) of a single diagnostic gene was greater than 0.75. The results of ingenuity pathway analysis (IPA) also indicated that DEGs were involved in the immune system and kidney disease-related pathways. Finally, we found multiple miRNAs that could regulate diagnostic genes from the ceRNA network. Conclusion. This study identified 6 IF/TA-related genes, which might be used as a new diagnosis model in the clinical practice.

Construction of Predictive Model of Interstitial Fibrosis and Tubular Atrophy (IFTA) After Kidney Transplantation with Machine Learning Algorithm

“…Furthermore, Gene scoring was used to eliminate batch effects in four data sets, which enabled our prediction model to show good predictive power in both the modeling and 3 validation groups. Our study shows that our model has a predictability (AUC = 0.956) superior to another IFTA forecasting model integrated COX regression model with LASSO constructed by Yang et al (AUC = 0.8210) (11). Moreover, the AUC of the predictive model achieved 0.938 in the validation set GSE53605, 0.781 in set GSE76882, and 0.720 in set GSE22459, which suggested that our model has a high level of accuracy without over tting.…”

“…In previous study of IFTA diagnosis based on mRNA expression, the quantity of samples and the method of ltering variables are limited, resulting in relatively low sensitivity and speci city (11). With the progress of machine learning algorithms, new approaches can be employed to diagnose diseases.…”

Downregulating ANGPTL3 by miR-144-3p promoted TGF-β1-induced renal interstitial fibrosis via activating PI3K/AKT signaling pathway

“…ANGPTL3's involvement in various renal pathologies, including diabetic nephropathy [ 6 ], renal cell carcinoma [ 7 ] and acute kidney injury [ 8 ] has been documented. From a bioinformatics perspective, Yang et al [ 9 ] identified ANGPTL3 as a crucial diagnostic gene connected to immune and renal disease-related pathways. This aligns with findings by Maluf et al [ 10 ], who in their genomic study observed a downregulation of angiogenesis-associated genes, including ANGPTL3, ANGPT2, and VEGF, in conditions of tubular atrophy and interstitial fibrosis.…”

Construction of predictive model of interstitial fibrosis and tubular atrophy after kidney transplantation with machine learning algorithms

“…In a previous study of IFTA diagnosis based on mRNA expression, the quantity of samples and method of filtering variables were limited (Yang et al, 2022). With the progress of machine learning algorithms, early linear models such as Least Absolute Shrinkage and Selection Operator (LASSO), ridge regression (Ridge), elastic net (Enet), stepwise generalized linear models (Stepglm), linear discriminant analysis (LDA), partial least squares regression (plsRglm), and Naive Bayes have gradually evolved to non-linear models such as support vector machines (SVM), generalized boosted models (glmboost), random forest, gradient boosting machines (GBM), Extreme Gradient Boosting (XGBoost), and artificial neural networks (ANNs) and there are now many options available for the construction of disease diagnostic models.…”