AI Signatures Link De Novo Lipogenesis to Chronic Active Antibody-mediated Rejection in Kidney Transplants

Abstract

Antibody-mediated rejection (ABMR) remains a major cause of kidney graft dysfunction despite advances in immunosuppression. Recent evidence suggests that metabolic reprogramming, particularly de novo lipogenesis, may contribute to chronic allograft injury. This study aimed to identify gene signatures associated with lipid metabolism that are linked to chronic active ABMR in renal transplant recipients.
Publicly available microarray datasets and clinical metadata were integrated for 100 biopsy specimens (55 ABMR and 45 controls). After normalization and batch correction, supervised machine-learning models—Random Forest, Support Vector Machine, and Convolutional Neural Network (CNN)—were trained to distinguish ABMR from control samples. Differentially expressed genes related to lipid metabolism pathways were identified and correlated with histopathologic and serologic parameters according to Banff diagnostic criteria.
The CNN model achieved an accuracy of 88% and an AUC-ROC of 0.92, outperforming SVM and RF classifiers. Identified lipid metabolism-related gene signatures showed significant associations with markers of immune activation and graft injury, implicating de novo lipogenesis in the pathogenesis of chronic active ABMR. Functional enrichment analyses further supported dysregulation of fatty-acid biosynthesis pathways.
Integrating transcriptomic profiling with artificial intelligence modeling uncovered lipid- related molecular patterns discriminating chronic active ABMR from stable grafts. These findings provide potential biomarkers and mechanistic insight into metabolic dysregulation underlying antibody-mediated graft injury, offering a foundation for future translational validation.