Abstract
With the rapid progression in computational science, methods that integrate computer graphics, pharmacology and structural biology to accelerate drug development have become an important subject in drug design. Among these methods, building of quantitative structure-activity relationship（QSAR）model and molecular docking simulation are areas that drug companies especially interested in. In this thesis, the relation- ship between molecular convex hull property, docking energy and activity has been studied by using 338 HIV protease inhibitors as material. With the use of conventional regression and Logistic Regression, we have found that although the linear regression model can not be built, it does exist a clustering relationship between our convex hull descriptor and activity. In addition, there is not an obvious relationship between docking energy and activity. In order to build a model between activity and docking energy, a more precise scoring function and structure -searching algorithm should be developed.