Exploring Perturbations in Peripheral B Cell Memory Subpopulations Early after Kidney Transplantation Using Unsupervised Machine Learning

Abstract

Background: B cells have a significant role in transplantation. We examined the distribution of memory subpopulations (MBCs) and naïve B cell (NBCs) phenotypes in patients soon after kidney transplantation. Unsupervised machine learning cluster analysis is used to determine the association between the cellular phenotypes and renal function. Methods: MBC subpopulations and NBCs from 47 stable renal transplant recipients were characterized by flow cytometry just before (T0) and 6 months after (T6) transplantation. T0 and T6 measurements were compared, and clusters of patients with similar cellular phenotypic profiles at T6 were identified. Two clusters, clusters 1 and 2, were formed, and the glomerular filtration rate was estimated (eGFR) for these clusters. Results: A significant increase in NBC frequency was observed between T0 and T6, with no statistically significant differences in the MBC subpopulations. Cluster 1 was characterized by a predominance of the NBC phenotype with a lower frequency of MBCs, whereas cluster 2 was characterized by a high frequency of MBCs and a lower frequency of NBCs. With regard to eGFR, cluster 1 showed a higher value compared to cluster 2. Conclusions: Transplanted kidney patients can be stratified into clusters based on the combination of heterogeneity of MBC phenotype, NBCs and eGFR using unsupervised machine learning.