 |
Projects
|
|
| Several projects relevant to data acquisition, storage, analysis and visualisation are currently underway which contribute to achieving the goals of the Visible Cell™. These projects are summarised as follows:
Whole Mammalian Cell Reconstruction Brad Marsh and colleagues have developed methods for preserving and imaging pancreatic beta cell architecture with unprecedented reliability, and have extended the application of these techniques from immortalised cell lines that can be manipulated in tissue culture, through to pancreatic islets of Langerhans isolated from rodents or humans that remain viable for transplantation. This technology combines methods for:
3D reconstructions generated from beta cells prepared in this way have provided powerful new insights into the spatial complexity of inter-organelle relationships in insulin-secreting cells. Unfortunately however, even relatively "large area" (e.g. ~4µm x 4µm) high-resolution 3D reconstructions computed from tilt series images of comparatively “thick” (300-400nm) slabs by electron microscope tomography (ET) still represent just a narrow cross-sectional slice through a cell, typically about 1% of an entire beta cell's volume. Clearly, a multi-scale/multi-resolution approach that utilises efficient although imprecise methods for spatially marking up large numbers of 3D cellular reconstructions is required to understand how the distribution and size of organelles changes with respect to time and different physiological conditions. Thanks to advances in software and acquisition hardware and software it is now possible to collect a high-resolution dual-axis tilt series in an electron microscope and reconstruct it into a 3D tomogram in a few hours. However, before valuable 3D information can be extracted from these tomograms, compartments of interest must be delimited and turned into a 3D model - a process called segmentation. Using the current process of manual segmentation (tracing compartments by hand) can take months per tomogram. Clearly, automated methods for the segmentation of large volume cellular tomograms are required if large volumes are to be annotated within a reasonable timeframe. Automated segmentation, once reliable parameters are established for routine use, will also reduce the amount of subjectivity introduced by a human operator. Single Particle Analysis (SPA) is a 3D structure determination process that has relatively low impact on the specimen in question. It involves the computational alignment and merging of large numbers of 2D projection images of individual molecules, recorded using high-resolution cryo-electron microscopy. For high-resolution studies, the SPA process usually begins with the capture of ~10,000 high-resolution 2D projection images of randomly orientated target molecules suspended in a thin, vitreous ice layer. These single particles are then aligned and classified according to their orientation to yield average 2D projections, representing defined views of the target molecule. Using these class sums, the angles relating them are then calculated, before computing the first molecular 3D reconstruction. Projection images of this 3D reconstruction are then used as references in iterative rounds of single particle alignment. It should be noted that the 3D reconstructions describes the full 3D structure, not just the 3D surface. 3D reconstructions determined by SPA are used to dock atomic resolution structures of subunits and sub-complexes solved by NMR, electron and X-ray crystallography techniques. Docking the merged molecular 3D reconstructions (SPA, NMR & X-ray) into the Visible Cell™, provides the link to which metadata can be tagged. A whole range of membrane proteins, macromolecular assemblies and viruses are currently being analysed. Membrane proteins comprise 25-40% of all proteins and conduct a myriad of finely tuned reactions in every cell. The clear rate-limiting step in structure determination of membrane proteins in particular, is the production of high quality crystals for electron and X-ray crystallography. The aim of this project is to develop a new, systematic, template-mediated membrane protein crystallization process using a lipid layer strategy. In recent years electron-crystallography of membrane proteins has undergone major developments allowing membrane proteins to be crystallized within a near-native lipid bilayer environment in order to obtain 2D crystals. 2D crystalisation is well suited for the crystallization of a wide range of membrane proteins including those that are small, hydrophobic, and not amenable to traditional methods of 3D crystal production. Functionalised monolayer technology coupled with improved protein expression techniques has streamlined crystal production and made it possible to collect data to determine not only protien structure but its detailed interplay with components of the lipid bilayer. Modelling for Inference/Hypothesis Generation With the advent of live cell video microscopy, new types of mathematical analyses and measurements are possible. Many of the real-time movies of cellular processes are visually very compelling, but analysis of changing quantities such as surface area and volume often show that there is more to the data than meets the eye. Using real-time cell imaging based on geometric modeling, predictions may be made about quantities such as ligand or solute concentration, pH, membrane tension, pressure differences and the amount of membrane trafficked. The models give a much greater appreciation of dynamics of the cellular processes imaged, and often lead to generation and experimental verification of new hypotheses. Edge Detection and Image Enhancement Advances in 3D electron microscopy and image processing are providing considerable improvements in the resolution of subcellular volumes, macromolecular assemblies and individual proteins. However, the recovery of high-frequency information from biological samples is hindered by specimen sensitivity to beam damage. Low-dose electron cryo-microscopy conditions afford reduced beam damage but typically yield images with reduced contrast and low signal-to-noise ratios. Hence techniques are required to enhance images by reducing noise while preserving edge detail. Once noise is reduced, edges may be more accurately detected and images segmented. Experimental Subcellular Localisation Determining the subcellular location of a protein is essential
to understanding its biochemical function. A cell is divided into
different cellular compartments and each compartment is associated
with a different range of biochemical processes; by localising a
protein to a specific compartment, or set of compartments, the cellular
role of the protein can be inferred. This information can provide
insight into the functions of hypothetical or novel proteins and
can provide a more specific organellar context in which to investigate
a particular protein. Publicly available subcellular localisation
data can be linked to the organellar structures defined in the Visible
Cell™. This will enable queries on federated databases to
support hypothesis-generation and hypothesis-testing. The results
of such queries can be viewed graphically within the Cell Visualiser
when appropriate. Access to these data in a graphical context-specific
manner will be a very useful tool for research and teaching. Image databases offer the possibility to present, integrate and search the vast amounts of data being created by high throughput cell imaging. However, current systems lack one key component: the ability to search the images themselves. Searching is typically based on keywords and tags created by the experimenter without any real reference to the images themselves except insofar as the experimenter sees them. What is required is to quantify and understand the information in the images in an unbiased way. Generation of morphological, textural and other image measures allows the comparison of images so that, for example, images exhibiting similar characteristics can be found automatically, or the statistical significance of changes under differing experimental conditions can be tested. We are constructing an information infrastructure that will integrate database technologies to handle genome-phenome data, including molecular sequences, structures, interactions and networks, results of computational modelling and simulation, images, data from array-based experimental research, automatic annotations and metadata. A framework known as the Cell Framework has been developed to host the data referred to using generic and domain-specific technologies for data management and integration, mass storage, indexing and retrieval, as well as XML and specialised mark-up languages. Particular attention has been paid to develop a framework that will facilitate rapid visualisation.
|
