Grid Computing in 3D Electron Microscopy Reconstruction

Grid Computing in 3D Electron Microscopy Reconstruction

J.R. Bilbao Castro (University of Almeria, Spain), I. Garcia Fernandez (University of Almeria, Spain) and J. Fernandez (University of Almeria, Spain)
DOI: 10.4018/978-1-60566-374-6.ch020
OnDemand PDF Download:
No Current Special Offers


Three-dimensional electron microscopy allows scientists to study biological specimens and to understand how they behave and interact with each other depending on their structural conformation. Electron microscopy projections of the specimens are taken from different angles and are processed to obtain a virtual three-dimensional reconstruction for further studies. Nevertheless, the whole reconstruction process, which is composed of many different subtasks from the microscope to the reconstructed volume, is not straightforward nor cheap in terms of computational costs. Different computing paradigms have been applied in order to overcome such high costs. While classic parallel computing using mainframes and clusters of workstations is usually enough for average requirements, there are some tasks which would fit better into a different computing paradigm – such as grid computing. Such tasks can be split up into a myriad of subtasks, which can then be run independently using as many computational resources as are available. This chapter explores two of these tasks present in a typical three-dimensional electron microscopy reconstruction process. In addition, important aspects like fault-tolerance are widely covered; given that the distributed nature of a grid infrastructure makes it inherently unstable and difficult to predict.
Chapter Preview


Studying the structure of biological specimens on different levels is of vital importance in understanding their functionality and interactions. Different techniques like X-ray crystallography, nuclear magnetic resonance and electron microscopy (EM) are used for such purposes.

EM relies on the transmission electron microscope to obtain projection images of the specimen under study (Williams & Carter, 2004). Such projections are somewhat similar to the well-known medical radiographs – but with X-rays being substituted by highly energetic electron beams. The projections of the specimen are processed, using complex computational algorithms, to create a three-dimensional (3D) representation of the specimen that can then be thoroughly studied. 3D EM is therefore the result of applying certain computing algorithms to the 2D projections obtained from the specimen under study through the electron microscope (Frank, 2006).

Depending on the nature of the specimen and the problem under consideration, different reconstruction techniques are used. For example, when studying specimens in the cellular range, the so-called electron tomography (ET) is used. On the other hand, specimens in the macromolecular domain are usually studied through electron crystallography and single particle techniques, depending on their structural features (like, for example, symmetry).

The objectives of the work presented here is to develop and test computational tools which cope with high computational demands existing at different steps of a 3D reconstruction process. In particular, grid computing will be the computational platform chosen to tackle two different tasks; parameters optimization in algebraic reconstruction algorithms and the application of such algebraic algorithms in the 3D ET field.

Key Terms in this Chapter

Data Replication: Data replication consists of keeping multiple physical copies of the same logical file. It is intended for fault-tolerance purposes and performance improvement on data access. Data replication is commonly used on grid computing.

EGEE: Enabling Grids for E-sciencE is a project funded by the European Commission which aims to promote grid technologies usage and implementation.

Electron Tomography: Technique to derive the three-dimensional structure of an unique specimen that is imaged in the electron microscope by tilting it around a single axis; this technique is similar to computerized axial tomography in medicine.

Storage Element: On grid infrastructures, a storage element is where users data is stored. It usually hides the complexity of different types of storage technologies (pools of disks, backup units, etc.) making user interaction easier.

gLite: The current middleware used for the EGEE grid. It joins, extends and enhances previously existing grid middleware initiatives like EDG and LCG.

Fault Tolerance: In computing, is the ability of a code to manage and recover from a situation which, untreated, would make to application to fail and/or produce unexpected resultss.

Three-Dimensional Reconstruction: The computational process to derive the three-dimensional structure of an object from its two-dimensional projections (i.e. as radiographs).

Three-Dimensional Electron Microscopy: Technique to elucidate the three-dimensional structure of biological specimens from projection images taken by electron microscopy.

Complete Chapter List

Search this Book: