Updated PIRCHE-T2 model improves dnDSA risk stratification in kidney transplant recipients

BackgroundDevelopment of de novo donor-specific antibodies (dnDSA) remains a key risk factor for antibody-mediated rejection and graft loss in kidney transplantation. The PIRCHE-T2 model estimates immunogenicity by predicting donor-derived HLA peptides presented via the recipient’s HLA class II molecules. A recent update to the model integrates a new neural network-based peptide-binding predictor “Frost”, which requires further testing of clinical performance.MethodsWe compared the predictive performance of the previous and updated PIRCHE-T2 models in two independent kidney transplant cohorts from Zurich (n = 1194) and Basel (n = 387). PIRCHE-T2 scores were assessed at total and locus-specific levels and analyzed in relation to dnDSA incidence using ROC curves, Kaplan-Meier, and Cox regression models.ResultsThe updated PIRCHE-T2 model generated lower and more condensed scores but improved dnDSA risk stratification across both cohorts. Higher scores remained associated with increased dnDSA risk. Notable improvements were observed for HLA-C scores. HLA-DQ also showed enhanced performance in one-mismatch subgroups and Cox models, while HLA-A improvements were primarily seen in the Basel cohort. Results for other loci remained similar between models, although HLA-DRB1 showed cohort-specific variation, highlighting the need for context-specific threshold refinement.ConclusionOur findings demonstrate that the updated PIRCHE-T2 model refines immunological risk stratification in kidney transplantation, offering improved performance for certain loci and patient subgroups. Its application may support more precise donor selection and individualized immunological assessment. Given observed cohort-specific differences, future work should focus on optimizing thresholds and validating the model across diverse populations to ensure broader clinical applicability.