Computer-Assisted Design of Benzoisoxazol derivatives Inhibitors of Bromodomain-containing Protein 4 (BRD4) with Favorable Pharmacokinetic Profile

Abstract

We performed a relation computed-aided design based on the structure of benzo[d]isoxazol derivatives inhibitors (BDIO) derivatives, new potent inhibitors of the BRD4 protein. By using in-situ modifications of the three dimensional (3D) models of BRD4-BDIOx complex (Protein Data Bank (PDB) entry code: 5Y8Z) were prepared for the training and validation sets compounds of 29 BDIOx with observed inhibitory potencies (). We first built a quantitative structure activity relationship (QSAR) model in the gas phase, linearly correlating the calculated enthalpies of the BRD4-BDIOx complex formation with (; = 0,80) first and then a superior QSAR model was brought forth, correlating computed relative Gibbs’ free energies of complexation and ( = -0.1205 + 6.9374 ; = 0.96) which was then validated by a 3D-QSAR pharmacophore generation model (PH4) ( = 0.996 + 0.0554 ; = 0.95). The structural information of the active conformation of the training set BDIOs from the models guided us in the design of a virtual combinatorial library (VCL) of 99 225 analogs. We then filtered the VCL by applying Lipinski’s rule-of-five, in order to identify new BDIOs drug likely analogs. The pharmacophore (PH4)-based screening retained 106 new and potent BDIOs with predicted inhibitory potencies up to 158 times more active than the most active traing set BDIO1 (). Finally, the predicted pharmacokinetic profiles of the best potent of these new analogs () were compared to current orally administered anticancer drugs. This computational approach, which combines molecular mechanics and the Poisson–Boltzmann (PB) implicit solvation theory, the pharmacophore model, the analysis of BRD4-BDIOs interaction energies, the in-silico screening of VCL compounds, and the inference of ADME properties resulted in a set of new suggested BRD4 inhibitors for the fight against CRPC.