Establishment And Validation Of A Predictive Model For Differential Rejection Diagnosis By Gene Expression Profiling (GEP) In Renal Allograft Biopsies Using Banff Human Organ Transplant (B-HOT) Gene Panel
1CareDx, Inc., Brisbane, CA, 2Arkana Laboratories, Little Rock, AR
Meeting: 2022 American Transplant Congress
Abstract number: 9013
Keywords: Biopsy, Gene expression, Kidney transplantation, Rejection
Topic: Clinical Science » Kidney » 34 - Kidney: Acute Cellular Rejection
Session Information
Session Name: Late Breaking: Clinical
Session Type: Rapid Fire Oral Abstract
Date: Sunday, June 5, 2022
Session Time: 3:30pm-5:00pm
Presentation Time: 3:30pm-3:40pm
Location: Hynes Room 313
*Purpose: Large scale GEP studies evaluating the performance of Banff-endorsed B-HOT gene panel for allograft rejection diagnosis have not been published. In this study, we tested this gene panel in approximate 1400 formalin-fixed paraffin-embedded (FFPE) renal allograft biopsy (Bx) specimens to develop and validate a multi-class predictive model for GEP-based molecular diagnosis of antibody-mediated rejection (ABMR), T cell-mediated rejection (TCMR) and mixed rejection (MXR).
*Methods: The discovery cohort included 1050 Bx (276 ABMR [187 C4d+, 89 C4d-], 342 TCMR, 76 MXR, 356 no rejection [including 191 native kidney Bx with/without inflammation]). The validation cohort included 345 Bx (56 ABMR [43 C4d+, 13 C4d-], 95 TCMR, 18 MXR, 176 no rejection [including 11 native kidney Bx]). Pathologic diagnoses were made by the consensus among 4 expert renal pathologists following the Banff 2019 classification. Total RNA samples extracted from tissue curls of each Bx were tested with B-HOT gene panel (758 endogenous genes and 12 reference genes) assays on the NanoString nCounter. Only samples passing quality control metrics of raw gene expression data underwent further analysis. The predictive diagnosis model, assigning a different probability score for each of 4 diagnosis groups in each Bx, was fitted in the discovery cohort by the lasso regularized regression model and 10-fold cross-validation procedure and then evaluated in the validation cohort. Molecular diagnosis of each Bx was made by the highest probability score.
*Results: The final multi-gene model achieved overall diagnosis accuracy of 84.6% in the discovery cohort and 79.7% in the validation cohort. Sensitivity and specificity for diagnosing rejection (regardless rejection type) are 93.7% and 89.9% (85.2% and 88.1%) in the discovery (validation) cohort. Sensitivity and specificity for diagnosing each of 3 rejection types are summarized in table 1.
| Discovery (%) | Validation (%) | ||
| ABMR | Sensitivity | 85.9±2.1 (C4d+: 85.0±2.6, C4d-: 87.6±3.5) | 80.4±5.3 (C4d+: 76.7±6.4, C4d-: 92.3±7.4) |
| Specificity | 93.4±0.9 | 95.2±1.3 | |
| TCMR | Sensitivity | 86.0±1.9 | 70.5±4.7 |
| Specificity | 91.9±1.0 | 90.8±1.8 | |
| MXR | Sensitivity | 48.7±5.7 | 44.4±11.7 |
| Specificity | 99.0±0.3 | 97.6±0.9 |
*Conclusions: Performed in a large set of FFPE renal allograft Bx specimens with diverse histologic findings, our study generated a GEP-based molecular diagnosis model with good accuracy of diagnosing ABMR and TCMR. Our model showed lower sensitivity for MXR, consistent with previously published similar GEP studies using an alternate platform, and taken together may reflect underlying molecular heterogeneity of histologically diagnosed MXR.
To cite this abstract in AMA style:
Zhang H, Haun R, Collin F, Cassol C, Coley S, Wilson J, Stone S, Qu K, Tian W, Agrawal N, Shekhtman G, Larsen C. Establishment And Validation Of A Predictive Model For Differential Rejection Diagnosis By Gene Expression Profiling (GEP) In Renal Allograft Biopsies Using Banff Human Organ Transplant (B-HOT) Gene Panel [abstract]. Am J Transplant. 2022; 22 (suppl 3). https://atcmeetingabstracts.com/abstract/establishment-and-validation-of-a-predictive-model-for-differential-rejection-diagnosis-by-gene-expression-profiling-gep-in-renal-allograft-biopsies-using-banff-human-organ-transplant-b-hot-gene-p/. Accessed November 21, 2024.« Back to 2022 American Transplant Congress