MRI: Acquisition of a GPU-Enabled Computer Cluster for Molecular Modeling Applications

Kansas State University, Manhattan KS

Investigators

Abstract

This award is supported by the Major Research Instrumentation (MRI), the Chemistry Research Instrumentation, and the Experimental Program to Stimulate Competitive Research (EPSCoR) programs. Professor Christine Aikens from Kansas State University and colleagues Paul Smith, Daniel Andresen, Bin Liu, and Jeffrey Comer are acquiring a graphic processing unit (GPU) computer cluster. In general, computer clusters are constructed with central processing units (CPU) as the core computational hardware. GPUs were subsequently developed as processing units and used to accelerate the video display on computer monitors. Computational scientists have found a combination of CPU/GPU systems are especially fast for certain computations, especially simulations. Software developed for computational simulations have reached the stage at which they can provide quantitative or semi-quantitative accuracy for a variety of molecular, biomolecular, nanoparticle, and materials modeling applications. These simulations provide needed atomic-level insight into chemical mechanisms and biological interactions that are difficult to study experimentally. This system provides for the computational needs of molecular modeling applications across chemistry, physics, engineering and biology. This system is used for research and training at Kansas State University. Researchers from across the state including those at community colleges, non-Ph.D. granting institutions, and minority-serving institutions also have access to this system and receive significant computing time. This computer cluster enhances research and education. It aids the modeling nanomaterials; understanding the fundamentals of electron-electron and electron-nuclear interactions. This cluster is also used in elucidating the factors involved in determining the selectivities of alcohol dehydration products on zeolite catalysts and carrying out protein-nanoparticle modeling to understand protein conformational transitions that sample relevant protein-surface configurations. The computer cluster is used in understanding denatured state ensembles in protein folding and understanding the role of protein aggregation which may be relevant to Alzheimer's disease.