Identification of core genes in programmed cell death during ischemia-reperfusion injury in kidney transplantation based on machine learning and experimental validation

Objective To identify key patterns of programmed cell death (PCD) and core genes during ischemia-reperfusion injury (IRI) in kidney transplantation.

Methods Kidney transplant datasets were obtained from gene expression database, and PCD-related differentially expressed genes were screened The non-negative matrix factorization algorithm was used to classify patients and analyze subtype-specific biological functions and key PCD patterns. Machine learning models combined with univariate Cox regression and Kaplan-Meier survival analysis were employed to identify core PCD genes during IRI in kidney transplantation and explore their correlation with key PCD patterns. A rat kidney transplant model was used to assess IRI severity through hematoxylin-eosin staining, serum creatinine (Scr), blood urea nitrogen (BUN), and Western blotting for key gene protein expression.

Results Fourteen PCD-related genes were identified. Patients were classified into metabolic (subtype 1) and inflammatory (subtype 2) subtypes. Subtype 2 activated four key PCD patterns: pyroptosis, necroptosis, apoptosis, and immunogenic cell death. The optimal model (XGBoost-CV:10 fold+Lasso-CV:10 fold) and survival analysis identified MCL1, BAG3, and RHOB as core PCD genes during IRI in kidney transplantation, which were broadly correlated with key PCD patterns. Experimental results showed that compared to the sham group, rats in the model group had more severe tubular injury, higher Scr and BUN levels, and increased BAG3, RHOB, and MCL1 protein expression (all P<0.001).

Conclusions These four PCD patterns are crucial in the pathogenesis of IRI in kidney transplant. MCL1, BAG3, and RHOB may serve as potential biomarkers and therapeutic targets for IRI in kidney transplantation.