Identification and Cross-Platform Validation of Sparse Molecular Classifiers for Antibody-Mediated and...

Molecular classifiers are a promising tool to refine the diagnosis of antibody-mediated rejection (ABMR) and T-cell-mediated rejection (TCMR) after kidney transplantation. Despite this potential, the integration of molecular classifiers in transplant clinics has been slow, in part because of the complexity of current assays and lack of a consensus platform. Herein, we aimed to develop and validate sparse molecular classifiers for ABMR and TCMR using allograft tissue.

In a discovery cohort of 224 kidney transplant biopsies, lasso regression was applied on microarray gene expression data to derive a molecular classifier for ABMR and TCMR, respectively.

A 2-gene classifier for ABMR (PLA1A, GNLY) and a 2-gene classifier for TCMR (IL12RB1, ARPC1B) were identified. External validation (n = 403 biopsies) demonstrated preserved diagnostic accuracy for ABMR (area under the receiver operating characteristic curve [ROC-AUC]: 0.80, 95% confidence interval [CI]: 0.75-0.85) and TCMR (ROC-AUC: 0.83, 95% CI: 0.77-0.89), with the possibility to discriminate between pure and mixed rejection phenotypes. Complementary to their diagnostic potential, the molecular classifiers associated with accelerated graft loss in a second validation cohort (n = 282 biopsies) and identified allografts at risk for failure with histological lesions that did not reach the Banff thresholds for rejection. The computational approach was further validated using the Banff Human Organ Transplant (B-HOT) gene panel in 2 independent biopsy cohorts that were analyzed on the Nanostring nCounter platform (n = 66 and n = 80, respectively).

Rigid variable selection strategies can yield sparse molecular classifiers for allograft rejection phenotypes with preserved accuracy and prognostic value across different molecular diagnostic platforms, which may facilitate their interpretation and clinical implementation.