Parallel Computation of Voxelised Protein Surfaces with OpenMP

Abstract

Precise representations of protein surfaces are extremely useful when studying protein interactions and properties. Given their simplicity and ability to represent geometrical and physicochemical properties of proteins, voxelised surface representations have received a lot of interest in bioinformatics and computational biology applications such as interaction interface prediction, ligand-binding pocket prediction and both protein--ligand and protein--protein docking. Computing voxelised surfaces for large proteins can be challenging, as space-demanding data structures with associated high computational costs are required. This paper presents a fast, OpenMP-based parallel algorithm for the computation of the voxelised representation of the three main protein surfaces (van der Waals, solvent-accessible and solvent-excluded) at high-resolutions. The solvent-excluded surface computation is based on a region-growing implementation of the approximate Euclidean Distance Transform algorithm with Hierarchical Queues. The geometrical relationship between the solvent-accessible and solvent-excluded surfaces allows us to obtain the latter very efficiently by computing distance map values only for a small subset of the overall voxels representing the protein. The algorithm computes the contribution to the overall outer surface for each atom in parallel. The proposed methodology was experimentally compared to two previous MPI-based parallel implementations showing overall better speedup and efficiency metrics as well as lower computation times.