| Poor pharmacokinetic properties (absorption, distribution, metabolism andexcretion, ADME) and toxicity is one of the main reasons for the failure of drugdevelopment, and thus the ADMET evaluations for drug candidates at the early stageof drug development is very important. The experimental measures for ADMET aretime-consuming and expensive, and therefore there is a tremendous need fordeveloping high quality and reliable in silico models to evaluate ADMET propertiesfor compounds in libraries or virtually synthesized with computers. The main purposeof this thesis is to build an integrated ADMET information system: PKKB(PharmacoKinetics Knowledge Base). By taking advantage of the data in PKKB wewill establish a set of simple rules to predict ADMET properties for early evaluationand pre-processing of large compound libraries.In the first part of the thesis, we have constructed PKKB, an integrated ADMETdatabase, which includes2D and3D chemical structures, pharmacological information,experimental or calculated physicochemical properties, and especially high-qualityexperimental ADMET data for drug and drug-like molecules. Now, PKKB has morethan10,000entries for experimental ADMET measurements. As far as we know,PKKB is currently the most extensive ADMET experimental information system forthe public. In addition, most of the data in the PKKB were used for the development ofADMET prediction models by our group.In the second part, we have established a set of simple rules to predict ADMETproperties. This set of rules can be used as guiding principles based on the structural features, and the compounds satisfying these rules may have better ADMET properties.The ADMET data for15935different compounds from the literature and publicdatabases were used in this study. We first calculated the15molecular descriptors thatare widely used in ADMET modeling, and then identified three important descriptorsby the principal component analysis (PCA). The three important descriptors are MW,ionization state and logD (pH=7.4). Meanwhile, we evaluated the simple rules basedon logP, TPSA, logS (pH=7.4) and logSw. Systematic search was used to determinethe most optimal combination of descriptors in the simple rules. |
