Many complex diseases such as cancer, cardiovascular disorders, and schizophrenia may be understood as failures in the functioning of nested hierarchies of biomolecular and cellular networks. These nested hierarchies control a range of processes including the differentiation and migration of cells, remodeling of extracellular matrices and tissues, and information encoding in neuronal subsystems. Washington University has established expertise in cutting edge imaging, molecular biology and genomic technologies synergistic with computational approaches such as machine learning and unraveling the principles of hierarchical organization and dynamics of complex systems. This collective expertise is being leveraged to develop new drugs, improve our ability to interpret sophisticated imaging data, understand how populations of neurons act collectively to accomplish complex tasks, and model the onset and progression of complex diseases as dynamical rewiring of hierarchical, multi-scale networks. Biological network analyses provide a rich set of tools for organizing and interpreting the vast quantities of data produced by state-of-the-art experimental protocols. The rapid advancement of computationally intensive research in these areas is outstripping the capabilities of CPU-based high performance computing (HPC) systems. This application would support the acquisition and integration of a large-scale IBM high performance cluster of Graphics Processor Units (GPUs) to be added as an upgrade to the existing IBM-designed Heterogeneous High Performance Computing environment to form a state-of-the-art hybrid computing capability. Such a resource is essential to match the growing need for high performance computing at Washington University and to support state of the art research software applications that are optimized for GPU computing. The acquisition and integration of a high performance GPU cluster will solve critical computing challenges that exist within Washington University's growing NIH research portfolio. The proposed state-of-the-art hybrid GPU/CPU computing capabilities will be deployed within the framework of a stable, productive and rapidly growing resource center. The addition of high-capacity GPU computing capabilities will allow critical calculation to be performed in hours instead of days and enable substantial increases in productivity for existing projects covering a broad range of application areas as well as enabling new research directions.
Showing the most recent 10 out of 17 publications
