Objective. Improve diagnosis of late cellular rejection of renal allograft using methods of intelligent data analysis.

Materials and methods. We conducted an observational, retrospective, one-center, analytical, comparative study in
2 groups by the “case—control” type. The first group (AR) included participants with histologically proved late cellular rejection of the transplant and the formation of chronic transplant dysfunction (22 patients). The second group (STA) consisted of the recipients with a stable transplant function receiving standard immunosuppression (22 patients). Flow cytometry with the peripheral blood lymphocytes phenotype determining was used as a method of assessing the recipients’ immune status.

Results. As a result, we found significant differences in the absolute count of the effector memory T-cells making 0.147 (0.115—0.260) · 109 cell/L in the AR group and 0.106 (0.067—0.136) · 109 cell/L in the STA group (Р=0.0167). The relative and absolute counts of myeloid dendritic cells were different, too: 0.65 (0.36—0.73) % vs 1.05 (0.67—1.4) % and 0.039 (0.028—0.056) · 109 cell/L vs 0.063 (0.049—0.076) · 109 cell/L, respectively (Р=0.0009, Р=0.003). The numbers of the plasmacytoid dendritic cells were different between the study groups, too: 0.0038 (0.0021—0.0054) · 109 cell/L vs 0,005 (0.0035—
0.007) · 109 cell/L for the absolute counts (Р=0.0414) and 0.055 (0.04—0.085) % vs 0.09 (0.05—0.12) % for the relative counts (Р=0.0197). In order to amplify the diagnostic capabilities of the developed biomarkers, we performed mathematical modeling basing on the classification tree method. The developed algorithm of the solution rules includes seven dichotomic split-nodes and eight terminal nodes. The diagnostic characteristics evaluation showed that developed algorithm had 88.64% accuracy in alloimmune conflict verification thus significantly increasing the diagnostic potential of individual biomarkers.

Conclusion. The obtained data showed that the developed algorithm of the classification tree decision rules using the relative numbers of myeloid and plasmacytoid dendritic cells for dichotomy had good diagnostic characteristics and could be used to verify the renal allograph late cellular rejection.