Assessing rejection-related disease in kidney transplant biopsies based on archetypal analysis of molecular phenotypes

• DOI: 10.1172/jci.insight.94197
• Publication Date: 2017-06-14T17:03:34Z
• Authors: Jeff Reeve; Georg A. Böhmig; Farsad Eskandary; Gunilla Einecke; Carmen Lefaucheur; Alexandre Loupy; Philip F. Halloran
• Keywords: 03 medical and health sciences; 0302 clinical medicine; 3. Good health

Conventional histologic diagnosis of rejection in kidney transplants has limited repeatability due to its inherent requirement for subjective assessment lesions, a rule-based system that does not acknowledge diagnostic uncertainty. Molecular phenotyping affords opportunities increased precision and improved disease classification address the limitations conventional systems quantify levels Microarray data from 1,208 transplant biopsies were collected prospectively 13 centers. Cross-validated classifier scores predicting presence antibody-mediated (ABMR), T cell-mediated (TCMR), 5 related lesions generated using supervised machine learning methods. These used as input archetypal analysis, an unsupervised method similar cluster examine distribution molecular phenotypes rejection. Six archetypes generated: no rejection, TCMR, 3 associated with ABMR (early-stage, fully developed, late-stage), mixed (TCMR plus early-stage ABMR). Each biopsy was assigned 6 scores, one each archetype, representing probabilistic based on rejection-related properties. Viewed clusters, existing Banff categories, but there 32% disagreement, much it probably reflecting "noise" current system. Graft survival lowest developed late-stage ABMR, better predicted by archetype than diagnoses. The results provide new standard recalibrating systems.