Title: Parallel Methods for Protein Coordinate Conversion
Proteins contain thousands to millions of atoms. Their positions can be represented using one of two methods: Cartesian or internal coordinates (bond lengths, angles, etc.). In molecular dynamics and modeling of proteins in different conformational states, it is often necessary to transform one coordinate system to another. In addition, since proteins change over time, any computation must be done over successive time frames, increasing the computational load. To lessen this load we have applied different parallel techniques to the protein conversion problem.
The Cartesian to internal coordinate translation computes bond distances, bond angles, and torsion angles for each time frame by using the protein chemical structure and atomic trajectories as inputs. This direction is easily parallelizable and we realized several orders of magnitude speed up using various parallel techniques including a GPU implementation.
The reverse direction, is used in molecular simulations for such tasks as fitting atomic structures to experimental data and protein engineering. This computation has inherent dependency in the data structures since bond lengths and angles are relative to neighboring atoms. Existing implementations walk over a protein structure in a serial fashion. This thesis presents the first fast parallel implementation of internal to Cartesian coordinates where substructures of the protein backbone are translated into their own Cartesian coordinate spaces and then combined using a reduction technique to find global Cartesian coordinates. We observed orders of magnitude speedup using parallel processing.
Prof. Miriam Leeser (advisor)
Prof. Ningfang Mi, ECE
Prof. Jaydeep Bardhan