The 2D and 3D QSAR of 30 compounds with type 1 diabetes inhibitors has been studied by using semi-empirical methods. The parametrization (PM6) method is employed as the basic set to optimize the derivatives using Spartan 14 and PaDEL v2.20 are used to calculate the chemical descriptors. To obtain a reliable QSAR model, the data set using the Kennard-Stone method to divide the derivatives into training set and test set comprising 21 and 9 compounds, respectively. An optimal model for the training set with significant statistical quality was established. The same model was further applied to the test set pIC50 of the 9 compounds. In the 2D-QSAR study, the MLR analysis produced 2 models, where the best one is model 2 with SEE = 0.3227; r^2 = 0.7409; r^2 adjusted = 0.6952; F = 16.20625. In the 3D-QSAR study, Atom-based fashion and pharmacophore-based fashion alignment were used. The results showed that CoMFA (uvepls) (q2 = 0.6897; r2 = 0.9999) have good stability and predictability. The internal validation indicated that CoMFA (uvepls) MIFs possess good predictive power than COMFA (ffdsel). The molecular docking study showed three (3) conventional hydrogen bonds with Arg97, Glu63, and Tyr159. Two carbon-hydrogen bonds with Ala69 and His70. MD simulation (1ns) analysis on the docked compound 17 assisted in the further exploration of the binding interactions. Some crucial interactions like pi-pi-T-shaped and amide-pi-stacked were identified. Hydrogen bond interactions with Arg97, Tyr7, and Trp167, respectively, bind more closely to the ligand. These results can offer useful insights.