Christoph Gorgulla
Max Planck Institute for Molecular Genetics
Germany
Title: Virtual flow: A flexible workflow system for virtual screening procedures on high-performance computers
Biography
Biography: Christoph Gorgulla
Abstract
A vast arsenal of computational tools has been developed in the past which are useful for virtual screening, from the preparation of ligand databases to the postprocessing of the screening results. At the same time computer clusters such as high-performance computers or supercomputers became more and more available, which employ distributed resource management systems (batch systems) in order to handle and organize the jobs of the users. In order to be able to carry out large-scale virtual screening procedures in a fully automated fashion on computer clusters which are managed by a batch system we have developed VirtualFlow, a workflow system for virtual screening related tasks. Its features are: Scaling without limits with respect to the CPUs with essentially no overhead, monitoring and controlling of the workflow during the runtime and automatic compression /decompression/ archieving of the input and ouput files. The VirtualFlow framework has been implemented in two applications. The first one, VFLP (VirtualFlow for Ligand Preparation), is specialized on the preparation of ligand databases, taking SMILES as input and converting them into ready-to-dock molecules in any desired target format. The second implementation, VFVS (VirtualFlow for Virtual Screening), is specialized on the virtual screening procedure itself. VFLP and VFVS can be seamlessly joined to carry out virtual screenings on the largest scales.