Bilbao-Castro J R, Marabini R, Sorzano C O S, García I, Carazo J M, Fernández J J
Dept. Arquitectura de Computadores y Electrónica, Universidad de Almería, 04120 Almería, Spain.
J Struct Biol. 2009 Jan;165(1):19-26. doi: 10.1016/j.jsb.2008.09.009. Epub 2008 Oct 8.
Three-dimensional electron microscopy allows direct visualization of biological macromolecules close to their native state. The high impact of this technique in the structural biology field is highly correlated with the development of new image processing algorithms. In order to achieve subnanometer resolution, the size and number of images involved in a three-dimensional reconstruction increase and so do computer requirements. New chips integrating multiple processors are hitting the market at a reduced cost. This high-integration, low-cost trend has just begun and is expected to bring real supercomputers to our laboratory desktops in the coming years. This paper proposes a parallel implementation of a computation-intensive algorithm for three-dimensional reconstruction, ART, that takes advantage of the computational power in modern multicore platforms. ART is a sophisticated iterative reconstruction algorithm that has turned out to be well suited for the conditions found in three-dimensional electron microscopy. In view of the performance obtained in this work, these modern platforms are expected to play an important role to face the future challenges in three-dimensional electron microscopy.
三维电子显微镜能够直接观察接近其天然状态的生物大分子。该技术在结构生物学领域的重大影响与新图像处理算法的发展高度相关。为了实现亚纳米分辨率，三维重建中涉及的图像尺寸和数量增加，计算机需求也随之增加。集成多个处理器的新型芯片正以降低的成本进入市场。这种高集成、低成本的趋势刚刚开始，预计在未来几年将把真正的超级计算机带到我们的实验室桌面。本文提出了一种用于三维重建的计算密集型算法——代数重建技术（ART）的并行实现方案，该方案利用了现代多核平台的计算能力。ART是一种复杂的迭代重建算法，已被证明非常适合三维电子显微镜的条件。鉴于这项工作中获得的性能，预计这些现代平台将在应对三维电子显微镜未来挑战方面发挥重要作用。
