文摘
A three-dimensional quantitative structure activity relationship (3D-QSAR) study was carried out on a dataset of 62 cinnamoyl derivatives as human immunodeficiency virus-1 integrase (HIV-1 IN) inhibitors using k-nearest neighbor molecular field analysis (kNN-MFA). QSAR models were developed using stepwise-forward selection (SWF), genetic algorithm (GA), and simulated annealing (SA) variable selection approaches. Selected QSAR models were validated internally [using cross-validated squared correlation coefficient (q 2)] and externally [using predicted squared correlation coefficient (pred_r 2) and Y randomization] to determine their predictive ability. SA-kNN-MFA model (q 2?=?0.7669, pred_r 2?=?0.7566) was considered as the best model due to its better predictive ability (higher pred_r 2 value) as compared to SWF-kNN-MFA (q 2?=?0.8956, pred_r 2?=?0.5905) and GA-kNN-MFA (q 2?=?0.6431, pred_r 2?=?0.7525) models in external validation. Steric and electrostatic field descriptors were favorable for HIV-1 IN inhibition activity in the best model. HIV-1 IN inhibition activity of an external dataset containing 22 compounds was predicted to check the applicability domain of developed models. Drug-likeness and in silico absorption, distribution, metabolism, excretion, and toxicity (ADME/T) studies were also performed to determine the pharmacokinetic and toxicity profile of dataset compounds. Subsequently, HIV-1 IN inhibition activity of 53 proposed compounds was predicted using developed models and their ADME/T properties were also calculated. Results of the present study suggested that 3D-QSAR model developed by SA-kNN-MFA method may be helpful to understand the structural requirement for HIV-1 IN inhibition and may support the designing of potent antiretroviral agents.