Assembly of Biomolecular Gigastructures and Visualization with the Vulkan Graphics API

Area of Science:

• Computational biology
• Structural biology
• Molecular modeling

Background:

• Visualizing large biomolecular structures (millions/billions of atoms) is challenging due to data size and hardware limitations.
• Existing methods struggle with automated building and real-time rendering of complex systems like viruses or cellular components.

Purpose of the Study:

• To develop novel algorithms for constructing and visualizing large-scale all-atom biomolecular models.
• To overcome the limitations of current hardware and software in handling gigastructure data.

Main Methods:

• Introduced a coarse-grained "pet representation" (1/10th size) with covalent bonds for efficient packing and molecular dynamics simulations.
• Developed a grid-based algorithm for generating multiple levels of detail (LODs) for interactive visualization.
• Utilized modular building blocks and instancing to reduce GPU memory usage for giant models.
• Employed the Vulkan graphics API and mmCIF format for performance and efficient storage.

Main Results:

• Successfully built all-atom models of SARS-CoV-2, HIV, and a 3.6 billion-atom presynaptic bouton.
• Demonstrated efficient packing optimization and realistic DNA/RNA conformations using pet molecules.
• Achieved interactive visualization of gigastructures with optimized LODs across various molecular graphics styles.
• Maximized performance on consumer hardware using Vulkan and efficient data storage with mmCIF.

Conclusions:

• The pet representation and LOD algorithms enable the creation and interactive exploration of unprecedentedly large biomolecular models.
• This approach significantly enhances the feasibility of visualizing and studying complex biological systems at the atomic level.
• The YASARA program and Petworld platform provide accessible tools for researchers to build, share, and explore these giant structures.