Development of a clinical prediction model for diabetic kidney disease with glucose and lipid metabolism disorders based on machine learning and bioinformatics technology.

OBJECTIVE

In this study, we investigated the internal relationship between the pathogenesis of diabetic kidney disease (DKD) and abnormal glucose and lipid metabolism to identify potential biomarkers for diagnosis and treatment and investigated the role of the immune microenvironment of glucose and lipid metabolism disorders in the occurrence and progression of DKD.

MATERIALS AND

METHODS

The chip datasets GSE104948 and GSE96804 from the Gene Expression Common Database (GEO) were merged using the "lima" and "sva" software packages in R Software (4.2.3), and the merged dataset was used as the validation set. The intersection between the differential genes of DKD and the glucose and lipid metabolism genes in the MSigDB database was identified, and a nomogram of the incidence risk of DKD was built using three machine learning methods, namely LASSO regression, support vector machine (SVM), and random forest (RF), to validate the accuracy of the prediction model. Immune scores were conducted using the unsupervised clustering method, and patients were divided into two subgroups. The two subgroups were screened for differential genes for enrichment analysis. The differential genes of patients diagnosed with DKD were clustered into two gene subgroups for co-expression analysis. In this study, we utilized the Cytoscape software to construct a network of interactions among key genes.

RESULTS

Using machine learning, a diagnostic model was developed with G6PC and HSD17B14 as key factors. Enrichment analysis and immune scoring demonstrated that the development of DKD was related to the imbalance in the microenvironment brought about by glucose lipid metabolism disorders.

CONCLUSIONS

G6PC and HSD17B14 may be potential biomarkers for DKD, and the established predictive model is more helpful in predicting the incidence of DKD.