Archetypal analysis of deceased donor kidneys: A molecular approach for posttransplant outcomes

Donor kidney tissue-based transcriptomics may represent a new dimension for the prediction of kidney transplant outcomes. In this prospective, single-center study, 276 kidneys from 174 deceased brain-death donors were assessed by microarrays to identify phenotypes of procurement biopsies. Molecular classifiers (extreme gradient boosting, logistic, and Poisson regression) with 10-fold cross-validation were employed to categorize donors based on clinical variables (age, body mass index, hypertension, expanded criteria donor kidney) and histologic scores (vascular fibrous intimal thickening, interstitial fibrosis, tubular atrophy, arteriolar hyaline thickening). Archetypal analysis and linear mixed model were applied to determine molecular phenotypes and their association with 1-year posttransplant estimated glomerular filtration rate (eGFR) in 234 donor kidneys. Three molecular archetypes were identified. The "ideal" archetype (median donor age 42 years, low Kidney Donor Risk Index [KDRI], minimal chronic histologic changes) was associated with the highest 1-year eGFR, whereas the "marginal" archetype (68 years, extensive chronic changes, high KDRI) was associated with the lowest one. The "intermediate" archetype yielded better 1-year eGFR despite donor profiles similar to the marginal group. Although KDRI predicted 1-year eGFR, adding molecular archetypes improved model performance (Akaike Information Criterion [AIC] 80.0 vs 83.7; P < .05). External validation in an independent data set (n = 174, GSE147451) confirmed the predictive value of the model. Molecular profiling of procurement biopsies may help to identify donor kidneys with higher posttransplant eGFR.