Computer-aided QSAR/QSPR Study On The Prediction Of Properties Of Organic Compounds

Rational drug discovery requires an early appraisal of all factors impacting on a possible drug candidate in the subsequent preclinical, clinical and commercial phases of drug development. The study of ADME (A: absorption, D: distribution, M: metabolism E: elimination) properties of drug is an important research field in drug discovery. Some computer models have been built for predicting ADME properties of compounds.Unfavorable ADME properties have been identified as a major cause of failure for candidate molecules in drug development. Consequently, it is necessary for us to predict ADME properties using computer models, with main aim to increase the success rate of lead compounds to drugs.In the first part of this thesis, several quantitative models for the prediction of boiling point (BP), molar volume (MV), octanol/water partition coefficient (logP) and aqueous solubility (logS) of alkylbenzenes were built. Each alkylbenzene was described by a simple set of six numeric codes derived from its molecular formula. With a set of six numbers as input descriptors, multiple linear regression (MLR) and nonlinear multivariable regression (NLMR) were applied to build the Quantitative Structure Activity/Property Relationship (QSAR/QSPR) models, respectively. These models show good prediction ability and the r² (square of correlation coefficient) of all these models are above 0.95. In addition, for the BP models, the root-mean-square (RMS) errors are less than 9°C; for theMV models, the RMS errors are less than 6 cm³ mol^-1; for the logP models, the RMS errors are less than 0.23; and for the logS models, the RMS errors are less than 0.33.In the second part of this thesis, we also built a MLR model for pKa (acid/base dissociation constant) of 180 carboxylic acids with phenyl group. Carboxylic acids with phenyl are a complex group of compounds that have attracted enormous attention in recent years because of their biological activities. In this work, a quantitative pka prediction model for carboxylic acids with phenyl group was developed by multiple linear regression (MLR) analysis with six molecular descriptors. 180 compounds are divided into two parts: training set (120) and test set (60). For the training set, r (correlation coefficient) is 0.873 and sd (standard deviation) is 0.34; For the test set, the r (correlation coefficient) is 0.849 and sd (standard deviation) is 0.28.In summary, these QSAR/QSPR models can be used for the prediction of these important properties of similar compounds or drugs.