Premium
A preliminary MTD‐PLS study for androgen receptor binding of steroid compounds
Author(s) -
Bora Alina,
Seclaman E.,
Kurunczi L.,
FunarTimofei Simona
Publication year - 2007
Publication title -
international journal of quantum chemistry
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.484
H-Index - 105
eISSN - 1097-461X
pISSN - 0020-7608
DOI - 10.1002/qua.21311
Subject(s) - chemistry , quantitative structure–activity relationship , steric effects , test set , van der waals force , androgen receptor , principal component analysis , protein data bank (rcsb pdb) , molecule , training set , computational chemistry , steroid , biological system , stereochemistry , artificial intelligence , computer science , organic chemistry , cancer , prostate cancer , hormone , biochemistry , medicine , biology
The relative binding affinities (RBA) of a series of 30 steroids for Human Androgen Receptor (AR) were used to initiate a MTD‐PLS study. The 3D structures of all the compounds were obtained through geometry optimization in the framework of AM1 semiempirical quantum chemical method. The MTD hypermolecule (HM) was constructed, superposing these structures on the AR‐bonded dihydrotestosterone (DHT) skeleton obtained from PDB (AR complex, ID 1I37). The parameters characterizing the HM vertices were collected using: AM1 charges, X log P fragmental values, calculated fragmental polarizabilities (from refractivities), volumes, and H‐bond parameters (Raevsky's thermodynamic originated scale). The resulted QSAR data matrix was submitted to PCA (Principal Component Analysis) and PLS (Projections in Latent Structures) procedure (SIMCA P 9.0); five compounds were selected as test set, and the remaining 25 molecules were used as training set. In the PLS procedure supplementary chemical information was introduced, i.e. the steric effect was always considered detrimental, and the hydrophobic and van der Waals interactions were imposed to be beneficial. The initial PLS model using the entire training set has the following characteristics: R 2 Y = 0.584, Q 2 = 0.344. Based on distances to the model criterions (DMODX and DMODY), five compounds were eliminated and the obtained final model had the following characteristics: R 2 Y = 0.891, Q 2 = 0.591. For this the external predictivity on the test set was unsatisfactory. A tentative explanation for these behaviors is the weak information content of the input QSAR matrix for the present series comparatively with other successful MTD‐PLS modeling published elsewhere. © 2007 Wiley Periodicals, Inc. Int J Quantum Chem, 2007