Back

Development of a Flexible Multiplex Urine RNA Assay for Detection and Differentiation of Kidney Allograft Injury

Keslar, K. S.; Xu, W.; Zmijewska, A. A.; Margeta, D.; Bromberg, J. S.; Friedewald, J. J.; Abecassis, M. M.; Newell, K. A.; Morrison, Y.; Bridges, N. D.; Heeger, P. S.; Fairchild, R. L.

2025-08-27 transplantation
10.1101/2025.08.25.25334374 medRxiv
Show abstract

Non-invasive approaches to detect kidney graft injury and distinguish donor-specific immune-mediated injury from other causes of inflammation are needed to guide recipient therapy while avoiding the morbidities of transplant biopsies. We used a multi-plex platform to interrogate RNA isolated from kidney transplant recipient urine to investigate gene expression patterns distinguishing grafts with no injury vs. ongoing injury and further differentiate acute T cell-mediated rejection (TCMR) from BK virus nephropathy (BKVN). As a training set we quantified expression of 796 immune function genes from 25 control recipients with stable graft function, 17 with biopsy-proven acute TCMR, and 13 with biopsy-proven BKVN. We identified a 20-gene signature that differentiated intragraft injury from grafts with stable function (area under the curve (AUC), 0.991) and a distinct 40-gene signature distinguishing acute TCMR from BKVN (AUC = 1.00). Validation in separate 118 urine RNA samples obtained at time of surveillance or for-cause biopsies from Clinical Trials in Organ Transplantation (CTOT)-08 and CTOT-19 studies showed AUC of 0.77 for the 20-gene injury signature and AUC of 0.79 for the 40-gene signature. Our results highlight the utility of this flexible, non-invasive biomarker platform for rapid detection and differentiation of immune processes causing ongoing kidney graft injury.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.