| Molecular structural characterization is one of the most important fundamental steps in fields of drug design and pharmaceutical evaluation. The molecular structural descriptors should reflect the structural information relating to the activities of drug, which determine the success of quantitative structure-activity relationship (QSAR) study. Pharmaceutical effects are deemed to derive from results of molecular-receptor interaction, including electrostatic, steric and hydrophobic interaction. For the homogeneous drugs that hold the same matrix, the substituent effects may change the electrostatic, steric and hydrophobic properties of the molecular, resulting in its activity changed. So study on the functions of the substituent in moleculars with the same matrix have great significance in QSAR research.The substituent descriptors have been constructed in this dissertation considering diverse properties and functions of substituents, including: (1) substituent physicochemical parameters (SPCP) that derive from physicochemical properties of substituent, describing contribution of the substituent in the electrostatic, steric and hydrophobic effects on drug molecular, thus having definite physicochemical meaning and easy interpretation and more information, (2) substituent surface potential field parameters (SSPF) that derive from the calculation of surface potential field of substituent by random sampling and 8 kinds pseudo atomic probes, describing the electrostatic, steric and hydrophobic potential field of the substituent on drug molecular, and (3) novel amino acid side chain surface potential field parameters (ASSPF) that reflect the functions of amino acid and that can depict the structure of peptide.The SPCP descriptors have been applied in the structural characterization and QSAR study in derivatives of phenylindole, phenylthio HEPT, cyclic ureas, substituted phenyl ethylamine, diarylimidazole, and benzoic acid. The variables are reduced using stepwise multiple regression (SMR) method for the training set, and statistical results indicate that R2 in the MLR models are 0.900,0.904,0.853,0.960,0.940(0.779) and 0.926, respectively, and R2cv in LOO cross validation are 0.662,0.8487,0.7233,0.901,0.9140 (0.6515) and 0.8477, respectively. To validate the predictive power of resulting models, external validation are performed with Q2ext values being 0.8974,0.9294,0.7617,0.7653,0.8036(0.7121) and 0.8866, respectively. The models obtained also shows that substituents in nitrogen atom of indole ring of phenylindole derivatives have large activity, while that with strong electronic effects in 1 and 3 position of indole ring and the structure with hydroxy group presenting in the 4 position of 2-phenyl ring may decrease activity. The substituents in 5-position of the thymine ring of phenylthio HEPT have more activity than that of phenylthio groups, and the weak electronic and hydrophobic effects of the substituents in 5-position can enhance the anti-viral activity. The activity of cyclic ureas is blocked by steric hindrance of ortho and para substituents and hydrophobicity of meta and para ones, the biological activity of substituted phenyl ethylamine derivatives may be improved by introducing the steric substituents and electronic supplying para ones into phenyl ring. The methyl in diarylimidazole derivatives is detrimental to activity but benefit to selectivity, and substituents in phenyl ring with less width at ortho position, strong electronic effects and large width at meta position, and great hydrophobic... |