Data Flow Management Of Virtual Screening In Large-Scale Parallel Tasks Management System

Virtual screening is about selecting in silico the best candidate drugs acting on a given target protein. With the development of Grid computing, scientists are apt to implement virtual screening on Grids environment. Large-scale Virtual screening involves mass of data, the ligands number in the millions.When implement virtual screening on Grids, scientists should query data, group data, upload data, invoke many docking jobs on Grids and download results manually. The goal of the large-scale parallel tasks management system based on CSGrid is the automatism of the virtual screening. The system divides the virtual screening application into many parallel tasks and executes them on different Grid nodes, so the management of the jobs flow and data flow is very important. This paper discusses how to manage the data flow of the virtual screening in the system and design rational data flow configuration policy, so the large-scale data flow will work together well with the jobs schedule to implement the virtual screening automatically and improve the efficiency of the Grids.The data flow management function contains three modules. One is the selection condition customization module. It provides a graphical editor, in which the user can customize the selection condition. One is the data query, data grouping and data transfer Web Service module. This module query data from the distributed ligands database, then the ligands collection is grouped by the number of the rotatable bonds and transferred to the Grid node which needs data. The other is the unified data management space, it provides a view of all data on the grid notes and the functions of data upload and download, it can also be used to fetch result data parallelly. This paper focuses on user experience and implementing skills of the three modules.