Design Compound Comparison and Clustering Techniques Based on Graphics Processing Units

A new drug to market usually costs a lot of time for the research and the development, and invests huge amount of money. After decoding the human genome, the molecular biology and the proteomic fields make a remarkable advance; people understand more clearly on the disease generation and the disease mechanism. Protein plays an important role in the diseasing factor. Some diseases come from the damage of genes, and some diseases need protein(s) as the transmission media, therefore there are many researches of drug design focusing on the inhibitor and activator of proteins. Computer-Aided Drug Design (CADD) becomes an emerging research field, and it is helpful to improve the efficiencies of drug design and development. According to the theoretical computation, CADD predicts the properties of physical chemistry, and visualize the 3D-structure of protein, ligand, or their combinations with the functions of computer graphics. In addition, CADD can compute the relation of energy variations between them, or study the pharmacophore of ligand, in order to design new drugs. CADD is a kind of approaches, named rational drug design. Rational drug design is based on structure, property, and mechanism. There are two major approaches: structure-based approach and ligand-based approach. Researchers use these two approaches to develop drugs depending on the different drug design strategies. Structure-based approach mainly uses techniques of docking and de novo design, and the ligand-based approach uses QSAR and Pharmacophore techniques mostly. However, the most important work is how to divide the known inhibitors with biological activities into the training set and testing set in order to build an efficient inhibitor screening model by using QSAR and Pharmacophore. The results will greatly affect the prediction ability of models. Unfortunately, there are no popular and acceptable guidelines for this work. Therefore, the goal of this project is to design a tool for dividing inhibitors into the training set and testing set. The main works of this project are to develop a GPU-based compound comparison technique and design a clustering method for inhibitors. List the details of works in this project below. A. Develop GPU-based compound comparison technique A-1 Build a database consists of huge existing compounds. A-1 Compound comparison techniques based on single/multiple GPUs. A-2 Propose a load-balancing technology among GPUs. A-3 Design a friendly user and service interface. B. Design a clustering method for inhibitors B-1 Propose a clustering method for compounds. B-2 Design and evaluate a strategy for dividing inhibitors into sets. B-3 Build Pharmacophore models and verifies the designed strategy. B-4 Design a friendly user and service interface.

Project ID:PB10408-5745
External Project ID:MOST104-2221-E182-051