Artificial Intelligence-Assisted Prediction of Blood Cells Regenerative Capacity in Patients with Diabetes Mellitus

Introduction: Peripheral blood mononuclear cell (PBMC) regenerative capacity is impaired in pre/diabetes mellitus patients (DMPs). In vitro evaluation of PBMCs pre/post vasculogenic conditioning (VC) and integrative assessment of classical clinical risk factors using complex machine learning-based (ML) methods may help to distinguish patients' immune cell regenerative potential. Methods: Eighteen DMPs diagnosed with myocardial infarction and sixteen controls were enrolled. Isolated PBMC regenerative capability was evaluated pre/post-VC using endothelial progenitor cell colony formation assay/unit (EPC-CFA/U) and flow cytometry analysis. The following classification ML algorithms were used for data analysis: support vector machine (SVM), logistic regression (LR), random forest (RF), decision tree (DT), and adaptive boosting (AB). Results: An in vitro EPC-CFA assay revealed that DMPs showed decreased fresh PBMC-derived definitive EPCs (DEPCs) compared with controls. Decreased colony formation was observed in patients with a reduced number of circulating EPCs and they did not appropriately respond to VC according to ML findings. Furthermore, proinflammatory M1 macrophages significantly increased (P = 0.03) in the majority of DMPs (non-responders) compared to that in controls, whereas VC shifted the phenotype toward M2 along with CD4 + T and Treg cell expansion. Clinical, pre/post culture cell phenotype, and EPC-CFU data integration and comprehensive ML analyses distinguished “non-responder”-associated distinct biomarkers, clinical parameters, such as elevated proinflammatory cells and systemic inflammation, and elderly patients from those of responders. Notably, SVM prediction accuracy was 100%; both LR and RF were 86%; and those of AB and DT were 71% and 43%, respectively. Conclusions: Applying comprehensive ML methods, as well as integrative analysis of the findings of in vitro assays, such as EPC-CFA; such as EPCs, M1/M2 macrophages, and T cells pre/post-VC; and clinical-laboratory parameters, enabling us to precisely evaluate regeneration capacity of impaired PBMNC functions. Moreover, this cell culture methodology can be used before cell therapy to define optimal dose and cell ratio for transplantation.