Parallel Processing Symposium, International (1999)
San Juan, Puerto Rico
Apr. 12, 1999 to Apr. 16, 1999
Robert E. Lynch , Purdue University
Dan C. Marinescu , Purdue University
Hong Lin , Purdue University
Timothy S. Baker , Purdue University
We present new parallel algorithms for 3D reconstruction of objects from 2D projections and their application for the determination of the structure of macromolecules from electron micrographs. A multi-resolution orientation determination algorithm uses a parallel search to determine the `best fit' of a given image with images in a reference database. The 3D reconstruction algorithm uses Cartesian coordinates and permits the reconstruction of objects that do not posses symmetries. The method decomposes a large linear system into a set of smaller systems that can be solved independently on different processors of a parallel computer or on a cluster of workstations. The paper outlines the reconstruction process and the data partitioning and load balancing issues pertinent to the parallel implementation of the algorithm; preliminary results obtained on a SGI Origin 2000 system are given.
Parallel processing, Structural biology, Electron microscopy, 3D reconstruction, Scientific computing.
T. S. Baker, D. C. Marinescu, R. E. Lynch and H. Lin, "Parallel Algorithms for 3D Reconstruction of Asymmetric Objects from Electron Micrographs," Parallel Processing Symposium, International(IPPS), San Juan, Puerto Rico, 1999, pp. 632.