Quantitative structure-activity relationship (QSAR) modelling study of some novel carboxamide series as new anti-tubercular agents

Abstract

At first, the structure of the compounds was accurately drawn and optimized using the Spartan 14 software at DFT level of theory with B3LYP/6-31G** basis set in a vacuum. The diverse chemometric descriptors were computed from the optimized structures using the PaDEL descriptors software, and the division of the dataset into training and test sets was done based on Kennard-Stone’s algorithm. Five (5) models were generated from the training set using genetic function approximation, and model 1 was chosen as the best due to its robust internal and external validation metrics (R2train = 0.8563, R2adjusted = 0.8185, PRESS = 3.5724, average R¯¯¯¯2m (LOO-train) = 0.6751, Q2cv = 0.7534, R2pred= 0.7543, R2test = 0.6993) which passed the model criteria of acceptability. 6-Bromo-N-(2-(4-bromophenoxy) ethyl)-2-ethylimidazo[1,2-a] pyridine-3-carboxamide (compound 13) was used as the structural template for the in silico design due to its high pMIC, and it is within the model’s chemical space.