With this award from the Major Research Instrumentation (MRI) program and the Chemistry Division, Professor Thomas F. Miller and colleagues John F. Brady, William A. Goddard, Zhen-Gang Wang and Niles A. Pierce from the California Institute of Technology will acquire a computer cluster with graphical processing units. The proposal will enhance research in a variety of areas characterized as soft matter behavior/simulations. The projects include investigations aimed at the rational design of nucleic acid, protein and enzyme systems, conformational dynamics of proteins and molecular motors, enzyme-catalyzed electron-transfer and hydrogen-transfer dynamics, trans-membrane signaling and transport processes, the nucleation of membrane adhesion, protein secretion across a cellular membrane, the formation of gels, the dynamics of ring-polymer mixtures, and polymer-based tissue engineering.
A computer cluster is a group of linked processors that work in concert to achieve vastly more computational power that individual computers. These are employed to investigate complex problems using computational methods based on theoretical models and programs. Such calculations, often used in conjunction with experimental data, allow chemists and biochemists to better understand many types of complex chemical and biological phenomenon. This resource will be used by students and faculty to develop the use of computer clusters based on graphical processing units (GPUs) rather than CPUs. This approach can speed up calculations and simulations enabling larger, more complex systems to be investigated.
We have acquired and fully deployed the Soft Matter Simulator (Somasim), a large parallel cluster of graphical processing units (GPUs). Somasim includes 32,256 GPU cores, 144 CPU cores with 912G DDR 3 memory, and a 72 TB storage array. It is fully functional and housed and maintained in a dedicated facility. It is the shared evenly among the six Caltech faculty involved in the grant. Assemblies of soft materials such as proteins, nucleic acids, polymers, colloids, gels, micelles, membranes, vesicles, emulsions, suspensions, liquid crystals, self-assembled monolayers, and Langmuir-Blodgett films play a fundamental role in life and industry, but there remain enormous challenges to understanding their properties and to designing improved systems. These challenges arise from the huge number of degrees of freedom in soft matter systems, which are characterized by weak molecular interactions between structural elements with a delicate balance between entropic and enthalpic contributions to the free energy. As the name implies, `soft' materials are easily deformed and driven far from equilibrium -- well beyond the linear response regime. Such materials exhibit complex and frustrated dynamics that span an enormous range of length and time scales, often over rugged free energy landscapes that are laden with metastable basins. To understand and predict soft matter behavior requires new tools and computational strategies. The investigators (Brady, Goddard, Miller, Mayo, Pierce and Wang) span the areas of biology, chemistry, engineering and applied mathematics, and they have utilized the Somasim GPU cluster to simulate and design of soft matter. Projects in these groups include the elucidation of transport across biological membranes, the design of nucleic acid regulatory pathways, the simulation of slow dynamics of colloidal glasses and gels, and the polymer-based engineering of cartilaginous tissue. In each of these research areas, the Somasim GPU cluster is spurring the development of new algorithms and large-scale parallelization strategies. The combined impact of the hardware and methods development is a dramatic advance in our ability to predict and design the properties of soft matter systems. As part of our educational outreach efforts, we have successfully implemented the TRIDENT program to bring high-school students and high-school teachers into our research group at Caltech for extended (6-week) research experience, which culminates in a presentation to graduate students and faculty. Additionally, numerous graduate students, post-doctoral, and undergraduate researchers have gained programming and computing experience on the Somasim machine via research projects and introductory tutorials. Completed projects in the research groups of the principle investigators include the elucidation of transport across biological membranes, the design of nucleic acid regulatory pathways, the simulation of slow dynamics of colloidal glasses and gels, and the polymer-based engineering of cartilaginous tissue. In each of ongoing research areas, the Somasim GPU cluster has spurred the development of new algorithms and large-scale parallelization strategies.