WGCNA与随机森林在CKD基因研究中的应用

WGCNA was used to cluster genes related to the occurrence and progression of CKD in GSE137570. Combining differential gene analysis, we identified 9 genes positively associated with CKD development and 20 genes negatively associated with CKD development. Using the random forest algorithm, the top 10 significantly important differential genes from the two subsets were selected. This resulted in three gene sets of different sizes: minimal (CCL2, SUCLG1, ACADM), medium (CCL2, GGT6, PCK2, SFXN2, SLC34A3, ALPL, GLTPD2, ACADM, SUCLG1), and maximal (CCL2, MMP7, GGT6, PCK2, SFXN2, SLC34A3, ALPL, GLTPD2, ACADM, SUCLG1). Evaluations of diagnostic performance using different scoring methods revealed that combinations of gene set size and scoring method had varying efficacies. The maximal plaque score achieved a diagnostic performance of 0.767 in GSE66494 and 0.760 in GSE180394. The medium z-score achieved a predictive performance for CKD progression of 0.687 in GSE60861. Cox regression analysis constructed a multivariate risk model with age, creatinine change, and medium z-score as variables. LASSO regression analysis built a multivariate risk prediction model with gender, age, medium z-score, and minimal ssGSEA as variables. Logistic regression analysis included only gender and eGFR before the observation period in the multivariate model. In the mouse UUO model, we found a high degree of similarity between mouse UUO and human CKD in KEGG enrichment, and the core genes related to the occurrence and progression of human CKD remained diagnostically valuable in mice.