Parallel High-Performance Computing Algorithm to Generate FEM-Compliant Volumetric Mesh Representations of Biomolecules at Atomic Scale

Abstract

The computational study of biomolecules has been undermined by the lack of models that accurately represent the structure of big complexes at the atomic level. In this work, we report the development of an algorithm to generate a volumetric mesh of a biomolecule, of any size and shape, based on its atomic structure. Our mesh generation tool leverages the octree algorithm properties with parallel high-performance computing techniques to produce a discretized hexahedral model faster than previous methods. The reported algorithm is memory efficient and generates volumetric meshes suitable to be used directly in Finite Element Analysis. We tested the algorithm by producing mesh models of different biomolecule types and complex size, and also performed numerical simulations for the largest case. The Finite Element results show that our mesh models reproduce experimental data.