A Java tool for dynamic web-based 3D ... - Semantic Scholar
Bioinformatics © Oxford University Press 2004; all rights reserved. Bioinformatics Advance Access published November 5, 2004
A Java tool for dynamic web-based 3D visualization of anatomy and overlapping gene or protein expression patterns
Victor E. Gerth1 and Peter D. Vize1,2*
Departments of Biological Science1 and Biochemistry & Molecular Biology2, University of Calgary. Alberta T2N 1N4, Canada
telephone:
403 220 8502
fax:
403 289 9311
e-mail:
[email protected]
* corresponding author
STRUCTURED ABSTRACT Summary: The Gene Expression Viewer is a web launched three dimensional visualization tool, tailored to comparing surface reconstructions of multi-channel image volumes generated by confocal microscopy or micro-CT. Availability: Software is available on request from the authors and is free to academics. Contact: [email protected] Supplementary information: Revisions and additional documentation to the Gene Expression Viewer are maintained at http://www.cbrbio.ucalgary.ca/3DModels and http://www.xenbase.org/3DModels. The idx3dIII Advanced Programming Interface is freely available for non-commercial use at http://www.idx3d.ch/idx3d/idx3d.html
Gene or protein expression data provides information on when and where genes are activated. In addition to indicating the site of action, this data is critical for confirming predicted protein-protein interactions, potential ligand-receptor associations, gene regulatory networks and developmental targets. The localised expression of a gene or protein can be detected using systems that stain expressing tissues with a dye or with a fluorescent tag that can be visualised using fluorescent microscopy. A laser scanning confocal microscope collects fluorescent data by raster scanning an optical slice and a pinhole is used to exclude data outside of the optical slice. Through incrementally acquiring a series, or stack, of optical slices in the Z plane 3D data sets can be created (Pawley, 1990). Within each plane data can be detected for a number of different channels, each reflecting the labelling of a target with a different form of fluor. Other imaging options are available, but at the present time confocal microscopy is the highest resolution and most widely used system (Ruffins et al., 2002). GENXL is a JAVA applet which launches transparently from a web browser (figure 1). It requires a server capable of using the hyper text transfer protocol to serve the application and geometry files. Three-dimensional projections are then software rendered on the client computer using a close variation of the IDX3DIII Application Programming Interface (API)(Walser, 1999). The IDX3DIII API is free for non-commercial use and includes all source code. Designed for JAVA 1.1 compliancy, GENXL works on all JAVA 1.1.3 compliant web browsers. Performance does vary with different browsers and operating systems and the fastest performance is currently achieved using Internet Explorer running on Windows. Two formulae are specific to GENXL. They illustrate our applied usage of integrated methods contained in the IDX3D III API source code. The algorithm for measuring the distance between two points on a surface is an approximation with an accuracy inversely proportional to the area of the two triangles coplanar with the points being measured. Distances are measured by applying the centroid of two triangles to a distance formula.
While the standard Euclidean metric L2 norm based on the Pythagorean theorem would be for an n-dimensional vector v=(v1, v2,..., v3), and two points P=(p1, p2,..., pn) and Q=(q1, q2,..., qn) the distance between them given as: QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
(Euclid, 300 BC; Sunday, 2001). We derive P and Q from a 3 dimensional vector produced by utilising an IDX3D III API method for acquiring the centre of a triangle, which is also the Median Point (Johnson, 1929). The integration of both subtractive and additive colour arithmetic for surface approximation required some restructuring of the scan line rendering process to support a material shader methodology. Calls to the IDX3DIII renderlineP() function have been replaced with the new function renderLineShader(shader) function, with shader being the index number for the render method. A surface with properties associated with the Phong (1975) method of shading implemented by renderlineP() can be supplemented with custom methods which implement variations of linear and non-linear algorithms (Kay, 1979). For additive transparency we have utilised: final = (background+colour) + (specular * reflectivity) where: background = the colour of the screen position colour = the colour of the surface specular = the specular value obtained from the light source reflectivity = a shininess value final = the final pixel value of the image Additive transparency allows overlaps between 3D expression domains to be easily visualised. In figure 1 this is illustrated by a screenshot depicting GENXL displaying two overlapping but distinct gene expression patterns. The first pattern appears green, the second red, and overlap is yellow. The application itself, including examples depicting immunofluorescence, in situ hybridisation and microCT data, can be
viewed at http://www.xenbase.org/3DModels/bioinformatics/. The Gene Expression Viewers' ability to simultaneously display, reorient, and organise by channel, multiple expression patterns provides a nearly instant method of comparing a multitude of 3D data varying by method of data acquisition, intended imaging objective, specimen, and anatomy. Effectively, this has allowed our application to contribute to 3D expression comparisons of genes predicted to be involved in early kidney development and comparison of their spatial distribution at varying stages in early development. Expansion of our database of 3D gene expression patterns will permit novel gene expression patterns to be retrieved through an indexed anatomical atlas by species, stage of development, gene, or protein. Our geometric projections are progressing towards the addressing of gene expression information in space and time in a comprehensive model, which in turn may allow computed virtual stages of development (where no volume data currently exists). We have begun expanding surface definition methods as the Gene Expression Viewer's current geometric projection techniques are functionally comparable to cutting edge hardware dependent implementations for three dimensional visualisation. The Gene Expression Viewer has already implemented experimental geometric visualisations utilising a spline based surface reconstruction (see supplemental data). This will permit geometry detail to be increased and decreased based on the capabilities of the client workstation. Our spline based implementations will be necessary for our approaches to warping dissimilar specimens and dynamic time based visualisations. The Gene Expression Viewer makes 3D visualisation instantaneously accessible over the internet to the largest segment of modern computers, adding a convenient tool for collaborative efforts between laboratories and disciplines . References Euclid, (300 BC) The Elements, Alexandria.
Kay, D. and Greenberg, D. (1979) Transparency for computer synthesized Images ACM SIGGRAPH Computer Graphics , Proceedings of the 6th annual conference on Computer graphics and interactive techniques,E 13:158-164. Pawley, J. B., ed. (1990). "Handbook of Biological Confocal Microscopy." Plenum, New York. Phong, B. (1975) Illumination for computer generated images. Communications of the ACM, 18(6), 311-317. Ruffins, S., Jacobs.R, and Fraser,S. (2002) Towards a Tralfamadorian view of the embryo: multidimensional imaging of development Current Opinion in Neurobiology 2002, 12:580-586. Sunday, D. (2001) http://geometryalgorithms.com/Archive/algorithm_0102/algorithm_0102.htm. Walser,P.(1999) http://www.idx3d.ch/idx3d/idx3d.html
Figure 1. GENXL displaying two different gene expression patterns in the developing embryonic kidney (http://www.xenbase.org/3DModels/bioinformatics/). The sodium glucose transporter is shown in light gray and the Na+K+ ATPase in dark gray. Regions
of overlap appear white (see above website for color versions). By activating the additive checkbox for channel 1 the overlap is easy to visualize, as it is by altering the opacity of one of the two channels with additive turned off. Filename kidney2.3DS.
A Java tool for dynamic web-based 3D visualization of anatomy and overlapping gene or protein expression patterns
Victor E. Gerth1 and Peter D. Vize1,2*
Departments of Biological Science1 and Biochemistry & Molecular Biology2, University of Calgary. Alberta T2N 1N4, Canada
telephone:
403 220 8502
fax:
403 289 9311
e-mail:
[email protected]
* corresponding author
STRUCTURED ABSTRACT Summary: The Gene Expression Viewer is a web launched three dimensional visualization tool, tailored to comparing surface reconstructions of multi-channel image volumes generated by confocal microscopy or micro-CT. Availability: Software is available on request from the authors and is free to academics. Contact: [email protected] Supplementary information: Revisions and additional documentation to the Gene Expression Viewer are maintained at http://www.cbrbio.ucalgary.ca/3DModels and http://www.xenbase.org/3DModels. The idx3dIII Advanced Programming Interface is freely available for non-commercial use at http://www.idx3d.ch/idx3d/idx3d.html
Gene or protein expression data provides information on when and where genes are activated. In addition to indicating the site of action, this data is critical for confirming predicted protein-protein interactions, potential ligand-receptor associations, gene regulatory networks and developmental targets. The localised expression of a gene or protein can be detected using systems that stain expressing tissues with a dye or with a fluorescent tag that can be visualised using fluorescent microscopy. A laser scanning confocal microscope collects fluorescent data by raster scanning an optical slice and a pinhole is used to exclude data outside of the optical slice. Through incrementally acquiring a series, or stack, of optical slices in the Z plane 3D data sets can be created (Pawley, 1990). Within each plane data can be detected for a number of different channels, each reflecting the labelling of a target with a different form of fluor. Other imaging options are available, but at the present time confocal microscopy is the highest resolution and most widely used system (Ruffins et al., 2002). GENXL is a JAVA applet which launches transparently from a web browser (figure 1). It requires a server capable of using the hyper text transfer protocol to serve the application and geometry files. Three-dimensional projections are then software rendered on the client computer using a close variation of the IDX3DIII Application Programming Interface (API)(Walser, 1999). The IDX3DIII API is free for non-commercial use and includes all source code. Designed for JAVA 1.1 compliancy, GENXL works on all JAVA 1.1.3 compliant web browsers. Performance does vary with different browsers and operating systems and the fastest performance is currently achieved using Internet Explorer running on Windows. Two formulae are specific to GENXL. They illustrate our applied usage of integrated methods contained in the IDX3D III API source code. The algorithm for measuring the distance between two points on a surface is an approximation with an accuracy inversely proportional to the area of the two triangles coplanar with the points being measured. Distances are measured by applying the centroid of two triangles to a distance formula.
While the standard Euclidean metric L2 norm based on the Pythagorean theorem would be for an n-dimensional vector v=(v1, v2,..., v3), and two points P=(p1, p2,..., pn) and Q=(q1, q2,..., qn) the distance between them given as: QuickTime™ and a TIFF (LZW) decompressor are needed to see this picture.
(Euclid, 300 BC; Sunday, 2001). We derive P and Q from a 3 dimensional vector produced by utilising an IDX3D III API method for acquiring the centre of a triangle, which is also the Median Point (Johnson, 1929). The integration of both subtractive and additive colour arithmetic for surface approximation required some restructuring of the scan line rendering process to support a material shader methodology. Calls to the IDX3DIII renderlineP() function have been replaced with the new function renderLineShader(shader) function, with shader being the index number for the render method. A surface with properties associated with the Phong (1975) method of shading implemented by renderlineP() can be supplemented with custom methods which implement variations of linear and non-linear algorithms (Kay, 1979). For additive transparency we have utilised: final = (background+colour) + (specular * reflectivity) where: background = the colour of the screen position colour = the colour of the surface specular = the specular value obtained from the light source reflectivity = a shininess value final = the final pixel value of the image Additive transparency allows overlaps between 3D expression domains to be easily visualised. In figure 1 this is illustrated by a screenshot depicting GENXL displaying two overlapping but distinct gene expression patterns. The first pattern appears green, the second red, and overlap is yellow. The application itself, including examples depicting immunofluorescence, in situ hybridisation and microCT data, can be
viewed at http://www.xenbase.org/3DModels/bioinformatics/. The Gene Expression Viewers' ability to simultaneously display, reorient, and organise by channel, multiple expression patterns provides a nearly instant method of comparing a multitude of 3D data varying by method of data acquisition, intended imaging objective, specimen, and anatomy. Effectively, this has allowed our application to contribute to 3D expression comparisons of genes predicted to be involved in early kidney development and comparison of their spatial distribution at varying stages in early development. Expansion of our database of 3D gene expression patterns will permit novel gene expression patterns to be retrieved through an indexed anatomical atlas by species, stage of development, gene, or protein. Our geometric projections are progressing towards the addressing of gene expression information in space and time in a comprehensive model, which in turn may allow computed virtual stages of development (where no volume data currently exists). We have begun expanding surface definition methods as the Gene Expression Viewer's current geometric projection techniques are functionally comparable to cutting edge hardware dependent implementations for three dimensional visualisation. The Gene Expression Viewer has already implemented experimental geometric visualisations utilising a spline based surface reconstruction (see supplemental data). This will permit geometry detail to be increased and decreased based on the capabilities of the client workstation. Our spline based implementations will be necessary for our approaches to warping dissimilar specimens and dynamic time based visualisations. The Gene Expression Viewer makes 3D visualisation instantaneously accessible over the internet to the largest segment of modern computers, adding a convenient tool for collaborative efforts between laboratories and disciplines . References Euclid, (300 BC) The Elements, Alexandria.
Kay, D. and Greenberg, D. (1979) Transparency for computer synthesized Images ACM SIGGRAPH Computer Graphics , Proceedings of the 6th annual conference on Computer graphics and interactive techniques,E 13:158-164. Pawley, J. B., ed. (1990). "Handbook of Biological Confocal Microscopy." Plenum, New York. Phong, B. (1975) Illumination for computer generated images. Communications of the ACM, 18(6), 311-317. Ruffins, S., Jacobs.R, and Fraser,S. (2002) Towards a Tralfamadorian view of the embryo: multidimensional imaging of development Current Opinion in Neurobiology 2002, 12:580-586. Sunday, D. (2001) http://geometryalgorithms.com/Archive/algorithm_0102/algorithm_0102.htm. Walser,P.(1999) http://www.idx3d.ch/idx3d/idx3d.html
Figure 1. GENXL displaying two different gene expression patterns in the developing embryonic kidney (http://www.xenbase.org/3DModels/bioinformatics/). The sodium glucose transporter is shown in light gray and the Na+K+ ATPase in dark gray. Regions
of overlap appear white (see above website for color versions). By activating the additive checkbox for channel 1 the overlap is easy to visualize, as it is by altering the opacity of one of the two channels with additive turned off. Filename kidney2.3DS.
