Cn3D: a new generation of three-dimensional molecular structure viewer

Christopher W.V. Hogue

Trends in Biochemical Sciences 1997 22: 314-316

(reproduced with permission)


In June of 1996, the National Center for Biotechnology Information (NCBI) released a new three-dimensional molecular structure viewer, Cn3D (pronounced 'see in three-dee') that can be used on Macs, PC's and Unix platforms. Cn3D is integrated with the Entrez [1] retrieval system, which includes WWW-based and network client software for retrieving and viewing biological information spanning bibliographical references, complete genomes, nucleic acid sequences, protein sequences and three-dimensional structures. At the time of writing Cn3D software has been downloaded over 16 000 times since June 1996 for viewing three-dimensional structures from the Molecular Modeling Database (MMDB) [2].



Cn3D design goals

Cn3D was deveoloped with two primary goals in mind. The first was to provide better and more reliable visualization of protein architecture and functionally important features, without requiring the user to have to burrow through the contents of PDB [3] files. The second was to design the software such that it can readily accommodate a wide range of new features to be added in the future.



Browsing structures without PDB files?

With previous structure viewers, the first step in looking at a structure was to obtain a PDB data file. Cn3D users have convenient access to the entire contents of the PDB database in the MMDB format used by Cn3D. These files use the ASN.1 data description language [4], the common language of all NCBI data, whether it might be sequence, bibliographic, taxonomical or now, three-dimensional structure data. The MMDB database is constantly updated as new entries appear on the Brookhaven PDB ftp site.

There are several mechanisms for retrieving and loading MMDB-formatted structure data. First, Cn3D has a simple, built-in mechanism for fetching and saving any structure over the Internet. Second, Cn3D can be launched with a three-dimensional structure from a WWW-browser like Netscape or Internet Explorer as a 'Helper Application' ( see Box 1 for instructions for configuring this feature). Starting from the WWW-Entrez system (Box 1), one can perform a structure query and arrive at an MMDB Structure Summary page. Alternatively, one can start with a particular structure from the Brookhaven Protein Data Bank 3DB Browser WWW site (Box 1), from which you may select the MMDB link to switch to the MMDB Structure Summary page for the same structure. This method for retrieving structure data will work at installations that are containted by an Internet firewall and that cannot use the fetch operation built into Cn3D or the Network version of Entrez.

Two other methods are available for access to MMDB formatted data: from the NCBI QUERY Email server (Box 1), which provides MMDB file in an ASCII text format for users who lack WWW access; plus the entire MMDB database and the Cn3D software can be also obtained via ftp (Box 1) from the NCBI.

Simple controls

Several pre-selected styles for renedering and coloring structures are available from the menu bar, similar to those found in RASMOL [5]. Cn3D also has a more detailed set of pop-in control panels, which provide enhanced rendering and labeling options, as well as controls for the three-dimensional display. These control panels are hidden until requested by the user.

Rotating, zooming and moving the structure can be performed by simply using the mouse. The Viewer control panel has controls to adjust the color, perspective and brightness of the three-dimensional image. It also contains a set of tape-recorder-like controls for playing animations from three-dimensional structure files containing multiple structures in NMR ensembles or while displaying multiple structure superpositions (Fig. 1).





Figure 1.
Silicon Graphics Cn3D 1.1 showing 1URE, the structure of intestinal fatty acid binding protein complexed with palmitate. (a) Cn3D showing the ensembel of 20 structures. The Render control panel was used to alter the default view to add amino acid sidechains drawn with Wireframe rendering and colored by Residue. (b) The first model in the NMR ensemble in same structure, obtained by pressing the rewind button '[|<]' on the Viewer Control panel. The perspective (p), brightness (b) and color (c) of the image can be altered with the [+] and [-] buttons on the left of the Viewer control panel.



Box 1. Cn3D software and 3D data access


Reliable data for smarter default images

Cn3D uses data from the MMDB, which is in a different format from PDB database entries. Converting files into MMDB format requires extensive validation, making the data more reliable for software that makes three-dimensional images. By contrast to PDB data files, MMDB data files have a consistent interpretation and Cn3D assumes that the description of atoms and bonds within the MMDB data file is correct and that no special exception handling is required for rogue data files. Thus when you load a classic protein such as Watson and Kendrew's myoglobin (1MBN) into Cn3D, you get an image that resembles the textbook picture, showing secondary structure together with the characteristic heme ring emphasized with a bright orange iron sphere and bound oxygen.

An example of the default Cn3D view is that of the structure of Gal4 with a 19-mer DNA double helix (1D66; Ref. 6), shown in Fig 2. This structure contains DNA, protein and Zn2+ ions. The default Cn3D image shows an alpha-carbon backbone trace of the protein colored by secondary structure, with the alpha-helices highlighted with hollow cylinders. To contrast with the protein, the complete structure of the DNA molecule is shown, colored with a CPK color scheme. The Zn2+ ions are shown as metallic spheres. Smarter default images mean that studying the contents of three-dimensional structure data becomes easier, as one can immediately see the biologically interesting contents of a structure database file.

Some of the new content added into MMDB data files includes the secondary structure definitions and domain information as used in the VAST structure-structure comparision [7] service. The VAST service (Box 1) allows one to retrieve a list of all structurally similar proteins to any protein in the MMDB database, and it also provides three-dimensional structure data files to view the superpositions of similar proteins in a variety of file formats.


Figure 2.
Macintosh Cn3D 1.0 showing the default view of the complex of the DNA-binding portion of 1D66, the structure of Gal4 with a 19-mer DNA double helix.





Cn3D on the inside

There are three layers of programming used in Cn3D: (1) the NCBI ASN.1 input/output (I/O) library, which reads ASN.1 MMDB files into memory and writes them back out into files; (2) a new molecular programming interface layer called MMDB-API, the middle layer of Cn3D, representing the molecular structure; this provides fast methods for moving among the hierarchy of molecules, residues, atoms and bonds in the memory of the computer; and (3) a new three-dimensional graphics interface called Viewer3D.

Of these, the most obvious is the Viewer3D three-dimensional graphics presentation layer. This layer is used to make pictures composed of 'three-dimensional graphics primitives': spheres, cylinders, lines, polygons and text. Viewer3D contains all the programming that tracks mouse movements, rotates, translate and zooms the three-dimensional image, and handles the division of the video display color palette into continuous shades for three-dimensional rendering.

One performance measure of three-dimensional graphics is the number of image updates (frames per second) while rotating the molecule. Viewer3D is slightly slower than RASMOL, but it it faster than MAGE [8,9]. For all these programs the underlying image update rate is fast enough to provide a very fluid rotation on the current crop of microprocessors.



Cn3D central dogma?

Molecules come into view in Cn3D after proceeding through the three layers, each performing a different transformation to the data. To borrow the obvious biological analogy, the MMDB data file is "transcribed' into an in-memory data structure, 'translated' into the MMDB-API molecular structure representation, and 'folded up' into a list of three-dimensional graphics primitives, which are 'transported' into the Viewer3D compartment where they undergo display an manipulation by the Cn3D user.

The layered approach to the software design allows for a maximum amount of re-use of the individual layers, and it provides a path for future improvements which are easily accommodated by changing one layer at a time.

Looking ahead to Cn3D 2.0

The currently available version of Cn3D is 1.1. Cn3D 2.0 is under development. Cn3D 1.1 has prototypes of some features that will be used more extensively in Cn3D 2.0, such as point-and-click selection, allowing the user to select residues, atoms or entire molecules. The mapping of mouse operations and movements can be altered allowing alternative mouse behaviors such as XY, YZ or XZ rotations. New features like three-dimensional distance measurements will be implemented using this mechanism in Cn3D 2.0. The largest improvement in Cn3D 2.0 will be an ability to alter the rendering, coloring and text associated with selected items directly in the Cn3D display and save this information as a 'Structure-Feature' file. This will allow Cn3D users to make illustrations and annotations directly in Cn3D without having to edit complicated text files. Thus scientist will be able to exchange information without becoming experts in the underlying file format. In addition, Cn3D 2.0 is incorporating a structure-superposition mode making it the preferred browser of VAST [7] structure-structure superpositions, even providing 'animated' views of protein structure evolution.



Acknowledgements

Cn3D would not be possible without the hard work of: S. Bryant, H. Ohkawa and J. Ostell for bringing the MMDB database into being; K. Sirotkin for the ASN.1 code generator; A. Smirnov and D. Vakatov for their implementation of Viewer3D with guidance from J. Kans; J. Epstein and B. Brylawski for the Network Entrez and WWW-Entrez servers. I would also like to thank T. Madej and J-F. Gibrat for VAST and T. Wolfsburg and A. Marchler-Bauer for reviewing this manuscript. Postdoctoral support for this work was provided by the Natural Sciences and Engineering Research Council of Canada (NSERC). Finally I would like to thank R. Sayle and D. Richardson and J. Richardson who pioneered the free distribution of high-performance cross-platform three-dimensional structure viewing software and its source code, a tradition Cn3D is proud to uphold.


References

  1. Schuler, G.D., Epstein, J.A., Ohkawa, H. and Kans J.A. (1996) Methods Enzymol. 266, 141-162.
  2. Hogue, C.W.V., Ohkawa, H. and Bryant, S.H. (1996) Trends Biochem. Sci. 21, 226-229.
  3. Bernstein, F.C., et al. (1977) J. Mol. Biol 112, 535-542.
  4. Rose, M.T. (1990) The Open Book, A Practical Perspective on OSI. pp. 227-322, Prentice-Hall. .
  5. Sayle, R.A., and Milner-White, E.J. (1995) Trends Biochem. Sci. 20, 374-376.
  6. Marmorstein, R., Carey, M., Ptashne, M., and Harrison, S.C. (1992) Nature 356, 408-414.
  7. Gibrat, J-F., Madej, T., and Bryant, S.H. (1996) Curr. Opin, Struct. Biol.6, 377-385.
  8. Richardson, D.C., and Richardson, J.S. (1992) Protein Sci. 1, 3-9 .
  9. Richardson, D.C., and Richardson, J.S. (1994) Trends Biochem. Sci. 19, 135-138.

[Mount Siani Hospital]
Christopher W.V. Hogue
Samuel Lunenfeld Research Institute
Mt Sinai Hospital, 600 University Ave,
Toronto, Ontario, Canada M5G 1X5
Email: hogue@mshri.on.ca


Created. 11 Aug 1997
NCBI Home Staff Papers