LIV-3D-QSAR model for estrogen receptor ligands

We have employed the Local Intersection Volume (LIV), a three-dimensional (3D) local shape descriptor, to study quantitative structure-activity relationships (QSAR) of 2-arylbenzothiophene analogs of raloxifene (1), a selective estrogen-receptor modulator (SERM), using the raloxifene bound conformation (PDB code: 1ERR) to build the structures of all ligands. The best LIV-3D-QSAR model obtained by a combined GA-PLS optimization, Model 1, was derived from RMS three-atom alignment using a training set of 44 compounds. The fit of the pIC(50) values, expressed by the squared correlation coefficient, R-2, was 0.78. After LOO-cv, a predictive squared correlation coefficient, Q(2), of 0.63 was obtained. Model 1 has three LIVs (1091, 1554, and 1654) with positive coefficients and three LIVs (597, 1463, and 1655) with negative coefficients. Four descriptors (LIVs) show excellent correspondence with pharmacophoric groups of the raloxifene series of compounds in accordance with SAR studies. Most interesting is the result of the prediction for 14 compounds (test set) used for external validation. The results provide the tools for predicting the pIC(50) values of related compounds and for the design and synthesis of new ER ligands.