Pharmacophore-based Target Prediction For The Polypharmacological Profiles Of Drugs And Parallelization Of3D Molecular Similarity Calculation Based On Distributed Computing Model

Drug discovery and development is extremely costly and risky. Traditional paradigm in drug discovery mainly focuses on developing exquisitely selective drugs to one specific target, The drugs developed under this strategy have exhibited lower clinical efficacy and higher toxicity in the past ten years. However, with advances in systems biology and polypharmacology, it brings a new paradigm in drug discovery and development:network pharmacology which has already shown significant advantages in tackling two problems in improving drug design-efficacy and lowering toxicity. Based on the previous researches of our research group, this thesis presents two parts of our work in network pharmacology: improving the accuracy of pharmacophore based target predictions and developing3D molecule similarity distributing computational model.Previously, our research group has developed a pharmacophore-based target identification method, named PharmMapper. However, PharmMapper still has limitations in the accuracy of drug targets prediction, we proposed a method with applying statistic treatment to enhance the enrichment of pharmacophore-based targets identification. Based on this method, we conducted a retrospective virtual screening containing1490drugs and673targets annotations to evaluate the performance of the improved PharmMapper. As a result, the test results indicate that its computation accuracy is indeed obviously improved and it also has been proved to have ability of predicting adverse drug reactions(ADRs) related targets.Molecular similarity is the cornerstone of many structure-related analyses in drug design and widely used in network pharmacology. However, it is very time consuming when the number of molecules required to be computed is large, which in turn limits its further application in large scale databases. To overcome this limitation, we parallelized SHAFTS by adapting the approach to distributed computing model within the famous BOINC framework and developed the drug design distributed computing platform, named Drug@home. During the runtime of the platform, it had facilitated hundreds of types of computing practices from different regions, indicating the potential computing power for computer aided drug design.