Diagnosis of T-cell-mediated kidney rejection in formalin-fixed, paraffin-embedded tissues using RNA-Seq-based machine learning algorithms.

Molecular diagnosis is being increasingly used in transplant pathology to render more objective and quantitative determinations that also provide mechanistic and prognostic insights. This study performed RNA-Seq on biopsies from kidneys with stable function (STA) and biopsies with classical findings of T-cell-mediated rejection (TCMR). Machine learning tools were used to develop prediction models for distinguishing TCMR and STA samples using the top genes identified by DSeq2. The prediction models were tested on 703 biopsies with Affymetrix chip gene expression profiles available in the public domain. Linear discriminant analysis predicted TCMR in 55 of 67 biopsies labeled TCMR, and 65 of 105 biopsies designated as antibody-mediated rejection. The random forest and support vector machine models showed comparable performance. These data illustrate the feasibility of using RNA-Seq for molecular diagnosis of TCMR in formalin-fixed tissue. Application of the derived diagnostic algorithms to publicly available data sets demonstrates frequent coexistence of TCMR in biopsies designated as antibody-mediated rejection. This underrecognition of TCMR in renal allograft biopsies has significant implications with respect to patient care.