Machine learning algorithm-based identification and verification of characteristic genes in acute kidney injury

BACKGROUND: Acute kidney injury is a common renal disease with high incidence and mortality. Early identification of high-risk acute renal injury patients following renal transplant could improve their prognosis, however, no biomarker exists for early detection. METHODS: The GSE139061 dataset was used to identify hub genes in 86 DEGs between acute kidney injury and control samples using three machine learning algorithms (LASSO, random forest, and support vector machine-recursive feature elimination). We used GSEA to identify the related signal pathways of six hub genes. Finally, we validated these potential biomarkers in an in vitro hypoxia/reoxygenation injury cell model using RT-qPCR. RESULTS: Six hub genes (MDFI, EHBP1L1, FBXW4, MDM4, RALYL, and ESM1) were identified as potentially predictive of an acute kidney injury. The expression of ESM1 and RALYL were markedly increased in control samples, while EHBP1L1, FBXW4, MDFI, and MDM4 were markedly increased in acute kidney injury samples. CONCLUSION: We screened six hub genes related to acute kidney injury using three machine learning algorithms and identified genes with potential diagnostic utility. The hub genes identified in this study might play a significant role in the pathophysiology and progression of AKI. As such, they might be useful for the early diagnosis of AKI and provide the possibility of improving the prognosis of AKI patients.