| Estrogen is estrogen receptor(ER) ligand molecule and has a significant impact onhuman reproductive system, musculoskeletal system and nervous system. Estrogen relateddiseases include reproductive system cancer, neurological disorders and obesity et.al, Whichare caused by interactions between estrogen and ERs. Now Development of selective estrogenreceptor modulators has been of great concern to researchers involved in chemistry andpharmacology of anticancer drugs over the years. In this work, a data set of82ER modulatorswith ERα and ERβ inhibitory activities was built, and quantitative structure-activityrelationship methods (QSAR) based on the two linear (Multiple linear regression, MLR,partial least squares regression, PLSR) and a nonlinear statistical methods (Bayesianregularized neural network, BRNN) was applied to investigate the potential relationship ofmolecular structural features related to the activity and selectivity of these modulators. ForERα and ERβ, the performances of the MLR and PLSR models are superior to the BRNNmodel, obtaining more reasonable statistical properties (ERα: for MLR, Rtr~2=0.72, Qte~2=0.63;for PLSR, Rtr~2=0.92, Qte~2=0.84. ERβ: for MLR, Rtr~2=0.75, Qte~2=0.75; for PLSR, Rtr~2=0.98,Qte~2=0.80). And also the MLR method is more powerful than other two methods forgenerating the subtype selectivity models, resulting in Rtr~2=0.74and Qte~2=0.80.Surlfex-dock was also used to explore the possible binding modes of the modulators. Theresults show that the8th substituent of the naphthalene or the quinoline plane and the spaceorientation of these two planes contribute to the subtype selectivity mostly. Multiple proteincrystals were used to explore the problems in docking and we find no certain linearrelationship between binding affinity and docking scores, Also the differences of proteins andligands handling methods weight greatly in the docking procedure, which should be givenenough concern in the future application. |
PDF Full Text Request