Machine learning-based identification of key lipid metabolism-related genes and associated ceRNA networks in NAFLD

Abstract

AbstractBackground:With the developmentof the economyand the improvementof people's living standards, the incidence of NAFLD has gradually increased, but the pathogenesis of NAFLD is still unclear. The pathogenesis of NAFLD may be related to the lipid metabolism disorder that has been widely recognized. Therefore, this study aims to provide a theoretical basis for the prevention and treatment of NAFLD to identify key fat metabolism-related genes in NAFLD patients.Methods:Two datasets, GSE126848 and GSE135251, were downloaded from the GEO database, and lipid metabolism-related gene sets were downloaded from the MsigDB database. We first used these two datasets and the lipid metabolism-related gene sets to identify and analyze how the lipid metabolism-related genes expressed differently in samples of patients with NAFLD and normal tissues. To evaluate whether the lipid metabolism-related genes were associated with the patients' disease , we analyzed nine machine learning algorithms, LASSO, LR, RF, XGBOOST, GBM, ANN, Adaboost, DT, and MNB, for the lipid metabolism-related differential genes, and then selected the optimal model based on the AUC values , and found that the LASSO model had the best prediction effect . So we used LASSO regression which was used as a diagnostic model to screen the characteristic genes, and the key genes of NAFLD were obtained , and then the immunoinfiltration analysis of the key genes was performed using the ssGSEA method , and the ceRNA network was constructed based on the key genes . Real-time fluorescence quantitative PCR (RT-qPCR ) was used to detect the expression levels of key genes in the blood of NAFLD patients and healthy volunteers.Results:30 distinct genes involved in lipid metabolism were identified and their functional enrichment was analyzed.Therefore, after analyzing nine machine learning algorithms, the predictive model with the highest AUC value was established based on LASSO analysis. Single-gene PCA, LR, and ROC were used to assess the validity of the predictive models, and finally, AJUBA, EDN2, EGR1, FMO1, and HPGDS were used as lead genes.Immune infiltration analysis revealed that the abundance of macrophages, CD8 memory T cells, and regulatory T cells was significantly reduced, and the abundance of CD56 natural killer cells was upregulated in NAFLD. This serves as a benchmark for immunometabolic interactions in NAFLD. Etiology of NAFLD. In addition, a ceRNA network based on key genes was constructed.rt-qPCR results showed that EGR1 and HPGDS levels were significantly reduced in NAFLD compared with controls, which was consistent with our analysis. In contrast to the results, the expression levels of AJUBA, EDN2 and fmo1 were significantly reduced.Conclusions:This study provides a deeper understanding of the molecular pathogenesis of NAFLD.We used five key genes to construct a diagnostic model that is not only related to lipid metabolism but also has a good predictive effect. The immune infiltration of the diagnostic genes was also analyzed and a ceRNA network was constructed. Thus, these five key genes may play an important role in the diagnosis and treatment of NAFLD.