NIH National Research Resources have proven to be an effective mechanism for both the development of new technologies to support biomedical research and broad dissemination of these technologies throughout the research community. Each Resource brings its own set of skills and tools to bear on problems of national importance. By working together we further improve our tools through cross-fertilization and sharing of expertise and instrumentation. We have formal ongoing collaborative projects with three other P41 Centers.
Collaboration 1: UIUC: Improved Molecular Modeling and Rendering
CISMM and the Resource for Macromolecular Modeling and Bioinformatics at the University of Illinois at Urbana-Champaign have shared code, ideas, and information informally for several years (for example, the UIUC Visual Molecular Dynamics (VMD) program forms the largest user group for the UNC Virtual-Reality Peripheral Network (VRPN) library, and exact fast sphere-rendering techniques developed by the UNC Resource are being used in the UIUC VMD program). Recently, our collaboration has become more formal, with UIUC being a primary user to test the scalability of UIUCs NAMD codes on UNC’s BASS GPU supercomputer.
Collaborators: Jim Phillips and John Stone, Senior Research Programmers, University of Illinois at Urbana-Champaign.
- Aim 1: Scaling GPU- and MPI-based NAMD. The NAMD scalable molecular dynamics code has been ported to run on multiple high-end graphics processor units (GPUs). UIUC is investigating ways to scale performance on the BASS Infiniband-connected system with 180 GPUs.
- Aim 2: Inexpensive, improved-quality haptic interface for VMD. Novint has recently released its Falcon 3D force-feedback interaction device. At $190, this device is quite affordable. UNC is working to provide a VRPN driver for this product and also a smoother haptic coupling model between VMD and force display.
Aim 3: Broad-Illumination rendering of molecules. The comprehension of unstructured, complex structures such as molecules benefits from a variety of rendering techniques. For cases such as the accessibility of docking sites to ligands, the understanding of pocket depths and their relationship to external structures is enhanced by the use of broad light sources, such as the sky on a cloudy day. We propose to use the UNC BASS supercomputer to produce downloadable cine loops of the structures in the Protein Data Bank (PDB) to provide a new and useful view on these molecules and to accelerate the rendering of such views in UIUC’s VMD. We are joined on this aim by David Banks, Associate Professor of EE&CS, UT/ORNL Joint Institute for Computational Sciences and Harvard NeuroDiscovery Center. This project came online in October 2009 and is described in detail (with downloadable 3D views) here.
Collaboration 2: UConn: Model Extraction and Rendering
The UNC Resource and the Center for Cell Analysis and Modeling (CCAM) at the National Resource for Cell Analysis and Modeling at the University of Connecticut Health Center began working together on training, dissemination, and sharing of data sets and algorithms when UNC presented a software workshop at the CCAM annual course in 2007. The collaborations revolve around the Uconn Virtual Cell simulation program.
Collaborators: Leslie M. Loew, Professor of Computer Science and Engineering, James C. Schaff, Assistant Professor, University of Connecticut Health Center.
Aim 1: Membrane and Cytoskeleton Modeling. One of the most time-consuming parts of simulation is the development of an accurate geometric model of the cell to be studied. This includes both the membranes (cell, nuclear, vesicle) that separate the cell into compartments and the cell cytoskeleton. We aim to provide semi-automatic model extraction and optimization tools to enable more rapid model construction.
- Aim 2: Quantitative Fluorescence Comparison. The final goal is to drive chemical simulations based on the difference between experimental images and simulated fluorescence images of the Virtual Cell outputs. An early goal is to use forward simulation to determine what fraction of light in each layer is caused by out-of-focus light to enable quantitative estimation of fluorophore densities.
- Aim 3: Multi-Chemical Volume Visualization. Understanding the causal relationships among three or more chemical concentrations throughout the volume of a cell can be confusing. We aim to apply multivariate volume visualization techniques to display the time-varying relationships among up to five chemical concentrations or molecular states in simulations.
Collaboration 3: Cornell: Improved 3D Microscopy Analysis and Visualization
The NIBIB-funded Developmental Resource for Biophysical Imaging Opto-Electronics (DRBIO) at Cornell has invented and developed new 4D multiphoton microscopy techniques for over twenty years. Currently they support 25 collaboration projects representing an extraordinary user base for CISMM technology. In the spring of 2009 we initiated this new collaboration with one of the pre-eminent centers in volumetric imaging in both in-vitro and in-vivo settings.
Collaborators: Warren Zipfel, Associate Professor of Biomedical Engineering, Cornell. Rebecca Williams, Research Scientist and Adjunct Assistant Professor, Biomedical Engineering, Cornell.
- Aim 1: Multi-Fluorophore Volume Visualization. Understanding relationships among two or more chemical concentrations throughout the volume of a cell or tissue can be extremely difficult. We aim to apply multivariate volume visualization techniques to display the relationships among up to five fluorophores.
- Aim 2: 3D Image Analysis. Tracking tube-like features in 3D (collagen fibers, chromosomes), 3D localization of diffraction-limited spots in PALM images, and tracking changes in shape over time (bone cell growth in the growth plate) in the presence of noise and motion are capabilities needed by DRBIO users.
- Aim 3: Stereo Display Station. DRBIO users have tried using commercial Volocity software for the 3D display of their data sets, but do not use it routinely. We propose to place a Simple Stereo System running ImageSurfer at the DRBIO and CCAM Resources and to train their staff and collaborators on its use.
This collaboration ties together the strengths of four Resources and will enable the development of new tools and visualizations for the already-large VMD/NAMD user population, for Virtual Cell users, for DRBIO users, and for PDB users. It pushes development of a number of existing CISMM tools and is a conduit through which to disseminate these tools to three new set of users. It also provides ties to two sets of simulation tools that can be used by users of our UNC Resource to further understand their data sets and to a new pool of multi-fluorophore 3D data sets for our algorithm development.