Molecular pathological diagnosis and clinical decision-making system based on mRNA expression profile and immunohischemistry in kidney transplantation

Introduction: Conventional pathological diagnosis of kidney allografts has inherent limitations. The Molecular Microscope Diagnostic System (MMDx) and molecular pathological models based on the B-HOT gene panel have not been extensively applied in clinical practice in China. In this study, mRNA expression profiles were detected by multiplex circular probe technology in FFPE samples obtained from kidney allograft biopsy. By integrating immune cell infiltration data with mRNA expression profiles, a predictive machine learning model was established to improve the diagnostic accuracy of kidney allograft biopsy.
Method: In the pilot study, mRNA expression profiling was performed on residual FFPE tissues from 10 patients using multiplex circular probe technology, and the results were compared with RNA sequencing data derived from paired fresh specimens to evaluate their consistency. Clinical and pathological data from patients who underwent indicated kidney allograft biopsies between May 2022 and December 2025 were retrospectively collected. Based on the Banff histological classification, patients were categorized into four groups: antibody-mediated rejection(AMR), T-cell-mediated rejection(TCMR), mixed rejection(MR), and no rejection(NR). mRNA expression profiles were determined in FFPE samples, and immune cell infiltration was evaluated and quantified. The top 20 significantly upregulated genes were identified in each group and integrated with immune cell infiltration data. Machine learning algorithms, including Lasso regression and XGBoost, were subsequently performed to establish novel molecular diagnostic models for kidney allograft injury.
Result: mRNA expression profiles detected by multiplex circular probe technology in FFPE specimens were highly consistent with those obtained by conventional RNA assays in fresh tissues(R =0.91). In the retrospective cohort study, a total of 149 patients were enrolled(n = 52 in AMR group, n = 35 in TCMR group, n = 30 in MR group, n = 32 cases in NR group). The XGBoost model is superior to the Lasso regression model. The AUC values of AMR vs other groups, TCMR vs other groups, MR vs other groups, and NR vs other groups can reach 0.883, 0.967, 0.858, and 0.997 respectively.
Conclusion: Multiplex circular probe technology enables the detection of mRNA expression profiles using FFPE samples, and its results are highly consistent with those of conventional RNA detection in fresh tissues, making it suitable for mRNA expression profile analysis of kidney transplant biopsy samples. The XGBoost molecular diagnostic model constructed using mRNA expression profiles and immune infiltrating cell information can improve the accuracy of pathological diagnosis of kidney allografts. Further validation of the model performance in this study requires multi-center and expanded samples.