MolVa: Workshop on Molecular Graphics and Visual Analysis of Molecular Data 2019

Permanent URI for this collection

https://diglib.eg.org/handle/10.2312/2632788

Browse

Browsing MolVa: Workshop on Molecular Graphics and Visual Analysis of Molecular Data 2019 by Subject "Computing methodologies"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    A Massively Parallel CUDA Algorithm to Compute and Visualize the Solvent Excluded Surface for Dynamic Molecular Data
    (The Eurographics Association, 2019) Schäfer, Marco; Krone, Michael; Byska, Jan and Krone, Michael and Sommer, Björn
    The interactive visualization of molecular surfaces can help users to understand the dynamic behavior of proteins in molecular dynamics simulations. These simulations play an important role in biochemical and pharmaceutical research, e.g. in drug design. The efficient calculation of molecular surfaces in a fast and memory-saving way is a challenging task. For example, to gain a detailed understanding of complex diseases like Alzheimer, conformational changes and spatial interactions between molecules have to be investigated. Molecular surfaces, such as Solvent Excluded Surfaces (SES), are instrumental for identifying structures such as tunnels or cavities that critically influence transport processes and docking events, which might induce enzymatic reactions. Therefore, we developed a highly parallelized algorithm that exploits the massive computing power of modern graphics hardware. Our analytical algorithm is suitable for the real-time computation of dynamic SES based on many time steps, as it runs interactively on a single consumer GPU for more than 20 k atoms.
  • Thumbnail Image
    Item
    QuickSES: A Library for Fast Computation of Solvent Excluded Surfaces
    (The Eurographics Association, 2019) Martinez, Xavier; Krone, Michael; Baaden, Marc; Byska, Jan and Krone, Michael and Sommer, Björn
    Recently, several fast methods to compute Solvent Excluded Surfaces (SES) on GPUs have been presented. While these published methods reportedly yield interesting and useful results, up to now no public, freely accessible implementation of a fast and opensource SES mesh computation method that runs on GPUs is available. Most molecular viewers, therefore, still use legacy CPU methods that run only on a single core, without GPU acceleration. In this paper, we present an in-depth explanation and a fully open-source CUDA implementation of the fast, grid-based computation method proposed by Hermosilla et al. [HKG*17]. Our library called QuickSES runs on GPUs and is distributed with a permissive license. It comes with a standalone program that reads Protein Data Bank (PDB) files and outputs a complete SES mesh as a Wavefront OBJ file. Alternatively it can directly be integrated in classical molecular viewers as shared library. We demonstrate the low memory consumption to enable execution on lower-end GPUs, and compare the runtime speed-up to available state-of-the-art tools.