Integration of ligand and structure-based virtual screening for identification of leading anabolic steroids

•We apply a QSAR model and a docking study in a heterogeneous data set of 269 AS.•The QSAR model for the AS explains structural features of the steroidal backbone.•Docking procedure predict the association of AS with the human androgen receptor.•14 steroids were identified as lead; the best was 7α-methylestr-4-en-3,17-dione. Parallel ligand- and structure-based virtual screenings of 269 steroids with anabolic activity evaluated in vivo were performed. The quantitative structure–activity relationship (QSAR) model expressed by selected descriptors as the octanol–water partition coefficient, the molar volume and the quantum mechanical calculated charge values on atoms C1, C2, C5, C9, C10, C14 and C17 of the steroid skeleton, expresses structural features of anabolic steroids (AS) contributing to the transport and steroid–receptor interaction. On the other hand, computational simulations of a candidate ligand binding to a receptor study (a “docking” procedure) predict the association of these AS with the human androgen receptor (AR). Fourteen compounds were identified as lead; the most potent was the 7α-methylestr-4-en-3, 17-dione. It was concluded that a good anabolic activity requires hydrogen bonding interactions between both Arg752 and Gln711 residues in the cycles A with O3 atom of the steroid and either Asn705 and Thr877 residues in the cycles D of steroid with O17 atom.