Exa-scalable interactive visual analysis for life & materials sciences – ExaViz

We develop software tools dedicated to interactive visualization and analysis of both experimental and simulation data. These tools are designed in order to withstand the massive and complex amounts of data generated by numerical simulations, with the additional challenge to put together an application that is capable to integrate a large variety of existing code.
By designing and integrating methods at the interface of virtual reality, scientific visualization and parallel simulation, our approach targets two scientific challenges : modeling an entire flu virion and simulating the GLIC receptor recently described in publications in the journals Nature and PNAS.
We propose an interdisciplinary approach to handle these issues through the design of an evolutive environment based on component-based programming, dedicated to scientific visualization and visual analytics. This software environment relies on new and efficient means for the construction and deployment of applications onto a parallel architecture. It is simultaneously dedicated to experimentalists and theoreticians, enabling an interactive visual analysis of phenomena that will enhance the cooperation and exchange between chemists, physicians and biologists. The attractivity and accessibility of this visual analysis, initially designed for execution in a desktop environment, will be reinforced by a portability onto an ensemble of recent immersive tools : web, mobile platforms and virtual reality centers.

The project is still in its early phase, but very encouraging preliminary results were already obtained. A first biological system, the FepA membrane protein, was studied. A demo video is associated to the corresponding publication :
moais.imag.fr/membres/matthieu.dreher/FvNanoICCS1080p.mp4
A prototype for a generic tool was developed to explore ideas in scientific visualization and is offered to the scientific community : unitymol.sourceforge.net.
The project has also enriched a database containing nuclear magnetic resonance (NMR) parameters of materials, and improved the simulation tools used for the prediction of NMR spectra.
Two international experts are involved in the project as external collaborators : Professor MSP Sansom at Oxford University (large scale simulations of viral capsides) and Professor T. Ertl at Stuttgart University (visual analytics). Many other scientists from various domains have indicated their interest for this project (in electron microscopy, SAXS, quantum chemistry, etc.).

The project did already allow a first convergence between the implicated scientific domains : biology, materials and computer science. This interdisciplinarity is reinforced and extended to the scientific community through events such as a workshop on visualization in biology, the organization of a visualization day and an upcoming international conference on molecular simulation and visualization.
Porting our developments into an immersive (virtual reality) context and to mobile platforms (tablets) is taking shape and reveals a high potential to enrich future applications and uses to be developed.

Research carried out in this project was already published in several scientific articles and presented at several conferences, spanning applications in biology (PLoS One) and materials science (Chem Mater) as well as more technical aspects such as scaling up of the application to hundreds of computers (Procedia ICCS). We are reaching out to a broad public and are in touch with different scientific communities related to this work (for ex. vulgarisation in Actualite Chimique ; a plenary lecture at the MiFoBio school).

New challenges in life and material sciences rely on the precise knowledge of structures and functions from nanometric down to the atomic length-scales. High performance computing is currently reaching the exascale and has the potential to face this major technical challenge. There is need for new tools to cope with the growing size of the underlying simulations and datasets. In Life Science, molecular dynamics simulations are central to the study of complex biological assemblies. The size of such simulations has grown exponentially, but post-processing, analysis, visualization and exploration of the generated data have stalled. Exascale simulations require new, scalable software tools to mine this data and extract the relevant biological information. In materials science, nuclear magnetic resonance (NMR) spectroscopy, combined with numerical simulations, has the potential to play a major role in the molecular-level understanding of solid-state compositions and architectures, to ultimately aid the design of innovative advanced materials. Making this major step forward will necessitate an interface for the visualization of (potentially periodic) molecular structures and NMR spectra, in order to build currently missing bridges between experiments, numerical models and structural information.
The ExaViz project aims at developing a framework for the interactive visual analysis of experimental and simulation data, and capable of handling the complex datasets generated by exascale simulations and the difficulties inherent to the integration of heterogeneous codes. By coupling virtual reality, scientific visualization and parallel simulation, we target two grand challenge applications: modeling a complete influenza virus and analyzing simulations of the GLIC receptor that we recently published in the journals Nature and PNAS.
We propose to gather interdisciplinary skills to design from component-based approaches a specific programming environment for scalable scientific visualization and visual analytics integrating new and efficient ways of building and deploying the applications on PC clusters. This framework, dedicated to experimentalists as well as theoreticians, will enable an interactive visual interpretation of phenomena to encourage cooperation and exchange between chemists, physicists and biologists. While the classical desktop is the first target for controlling the analysis process, we will also support and experiment immersive environments such as CAVEs as well as web and mobile devices.
These objectives will be achieved though the unique combination of highly complementary expertise from five partners. The LBT has a long experience in molecular simulation of biological macromolecules and in virtual reality approaches. The CEMHTI partner provides know-how in the combination of experimental and numerical approaches to solid-state NMR spectroscopy for materials science. The LIFO and Moais partners have developed FlowVR, a component oriented middleware for running parallel interactive applications on PC clusters. They both have a long-standing experience in parallel algorithms, code coupling and interactive visualization. The LIMSI partner is at the forefront of state-of-the-art virtual reality platforms in Europe. Further support will finally be provided by external collaborations with Prof. M.S.P. Sansom (Oxford Univ.) for simulations of viral capsids running on thousands of processors, and Prof. T. Ertl (Univ. of Stuttgart), an international leader on rendering and visual analytics, for the integration of codes (eg Megamol) developed in his group.