The search for new parameters for the prediction of type 2 diabetes mellitus (T2DM) or its harmful consequences remains an important field of study. Depending on the low-grade inflammatory nature of diabetes, we investigated three proteins in T2DM patients: 1-aminocyclopropane-1-carboxylate synthase (ACCS), granulocyte-colony-stimulating factor (G-CSF), and Sma Mothers Against Decapentaplegic homolog-4 (SMAD4). In brief, sixty T2DM and thirty healthy controls had their serum levels of ACCS, G-CSF, SMAD4, and insulin tested using the ELISA method. The insulin resistance (IR) parameter (HOMA2IR), beta-cell function percentage (HOMA2%B), and insulin sensitivity (HOMA2%S) were all determined by the Homeostasis Model Assessment-2 (HOMA2) calculator. The predictability of these protein levels was investigated by neural network (NN) analysis and was associated with measures of IR. Based on the results, ACCS, G-CSF, and SMAD4 increased significantly in the T2DM group compared with the controls. Their levels depend on IR status and inflammation. The multivariate GLM indicated the independence of the levels of these proteins on the covariates or drugs taken. The receiver operating characteristic area under the curve (AUC) for the prediction of T2DM using NN analysis is 0.902, with a sensitivity of 71.4% and a specificity of 93.8%. The network predicts T2DM well with predicted pseudoprobabilities over 0.5. The model's predictive capability (normalized importance) revealed that ACCS is the best model (100%) for the prediction of T2DM, followed by G-CSF (75.5%) and SMAD4 (69.6%). It can be concluded that ACCS, G-CSF, and SMAD4 are important proteins in T2DM prediction, and their increase is associated with the presence of inflammation.
Read full abstract