Bioinformatics Advance Access originally published online on July 26, 2005
Bioinformatics 2005 21(17):3578-3579; doi:10.1093/bioinformatics/bti576
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Automated Microarray Image Analysis Toolbox for MATLAB
1Statistical Sciences, Pacific Northwest National Laboratory Richland, WA 99352, USA
2Biochip Technology Center, Argonne National Laboratory Argonne, IL 60439, USA
*To whom correspondence should be addressed.
 |
Abstract |
|---|
 TOP Abstract INTRODUCTIONSOFTWARE IMPLEMENTATIONCONCLUSIONSREFERENCES |
|---|
Summary: The Automated Microarray Image Analysis (AMIA) Toolbox for MATLAB is a flexible, open-source, microarray image analysis tool that allows the user to customize analyses of microarray image sets. This tool provides several methods to identify and quantify spot statistics, as well as extensive diagnostic statistics and images to evaluate data quality and array processing. The open, modular nature of AMIA provides access to implementation details and encourages modification and extension of AMIA's capabilities.
Availability: The AMIA Toolbox is freely available at http://www.pnl.gov/statistics/amia. The AMIA Toolbox requires MATLAB 6.5 (R13) (MathWorks, Inc. Natick, MA), as well as the Statistics Toolbox 4.1 and Image Processing Toolbox 4.1 for MATLAB or more recentversions.
Contact: amanda.white{at}pnl.gov
|
INTRODUCTION |
|---|
|
TOPAbstractINTRODUCTIONSOFTWARE IMPLEMENTATIONCONCLUSIONSREFERENCES |
|---|
Microarray technology is a powerful tool for simultaneously performing multiple genomic or proteomic assays on a sample. However, its usefulness can be limited by the software for microarray image analysis. A variety of proprietary and free microarray image analysis tools are available; however, these are often very restrictive with regard to the types of images that can be analyzed, the statistics that can be computed and the implementation details that can be gleaned. Proprietary software, such as ScanArray Express (PerkinElmer Life and Analytical Sciences, Inc, Boston, MA) is not open-source nor easily customized when an alternative analysis may be required. TIGR-Spotfinder (Saeed et al., 2003) is a free software, but requires paired 16-bit TIFF images as input. F-SCAN and P-SCAN (Carlisle et al., 2000) are free tools developed at NIH. F-SCAN is also designed specifically for two-color experiments, and P-SCAN only accepts images from Fuji and molecular dynamics scanners. Spot (http://spot.cmis.csiro.au/spot/index.php) is a proprietary but open-source image analysis tool designed for two-color experiments. Our research found no tools currently available that are free, open-source and applicable to single-channel microarray experiments. Critical diagnostics about the quality of the imagery, spot statistics and experimental procedure, although becoming more common, are not typically emphasized.
|
SOFTWARE IMPLEMENTATION |
|---|
|
TOPAbstractINTRODUCTIONSOFTWARE IMPLEMENTATIONCONCLUSIONSREFERENCES |
|---|
The AMIA Toolbox for MATLAB was developed to address these issues. AMIA is an open-source, open-development tool for use with MATLAB. AMIA is transparent and can be easily extended or modified. AMIA is distributed in the standard MATLAB language, so users can view and alter the algorithms, and add or substitute modules to perform different functions. This openness and flexibility makes AMIA ideal for those who want customized solutions for their microarray image analyses.
AMIA conveniently analyzes a large collection of microarray images, one image at a time. Very little user input is needed, including images and slide layout information, such as number and arrangement of arrays and spots within arrays, and location of positional markers within the arrays. It is easily configured to handle almost any microarray layout. The tool supports our work with low spot-frequency and custom arrays usually lacking the quality of commercial arrays, but AMIA also can be used with large, commercial arrays.
The software sequentially analyzes each image in a set of microarray images that have the same array layout. It elicits minimal user input to identify the layout for an entire image set through the display of one image in the set. AMIA then automatically locates the expected spot centers on subsequent arrays. Since spot locations and shapes may vary from those expected, AMIA optionally uses several methods to identify the pixels associated with each spot. The first assumes that each spot is shaped identically and the spot centers fit the expected grid. This spot location method, though crude, provides a starting point for more sophisticated algorithms. The second assumes that each spot is identically shaped but allows the center to vary within a small neighborhood from the expected position. In both methods, the expected spot shape is dynamically and empirically estimated from AMIA-selected spots within each image to allow for consistent imperfections in spot shape, say owing to a dragging print tip. The third method uses a seeded-region-growing algorithm (Hojjatoleslami and Kittler, 1998) to identify contiguous pixels that differ statistically from the background and allows the spot to assume a variety of testable shapes within a small neighborhood. This method is useful if spots vary significantly in location, size, shape or texture.
AMIA outputs extensive summary statistics for each spot and its neighboring background for each of the three methods. These summary statistics include mean, median, standard deviation and pixel count for spot and background, and spot shape statistics for the seeded region grower spots. The output also includes background estimates for each of the quadrants neighboring a spot. These statistics enable the user to test and adapt if a scratch or other slide imperfection in one quadrant is affecting the overall background estimate.
AMIA provides diagnostics about the performance of the statistical algorithms and diagnostics that highlight potential problems with the microarray imagery. Algorithm diagnostics include a variety of array image overlays displaying the microarray with the expected grid, or detected spot locations and perimeters for each of the three spot identification methods. These easily interpretable diagnostic visualizations help users understand algorithm performance, and unusual results, and select the appropriate summary statistics to take forward.
Since microarray processing—probe preparation, array printing, sample preparation, hybridization, imaging and spot statistics extraction—can contribute significant variability to the resulting spot estimates, AMIA also produces tabular and visual diagnostics to illuminate processing-related artifacts. For example, the correlation estimates of spot intensity and background intensity with slide position uncover slide printing problems or uneven lighting during imaging. Another processing artifact that may be introduced by CCD imagers is spot overshine, an extensive bright halo from bright spots, owing to the fact that all pixels in the image are illuminated and captured at once. When a multipixel region is illuminated, photons from neighboring pixels are captured when measuring the pixel of interest. Hence, a bright pixel shines over into its neighbors. (We do not see this in scanning laser imagers where only a single pixel is illuminated and measured at a time.) However, in CCD images spot overshine affects not only the background estimate for that spot but also the estimates of neighboring spots. To diagnose overshine, AMIA produces a spot by background scatter plot annotated with regression results. A goal of AMIA is to raise user quality awareness and encourage users to implement more useful diagnostics.
AMIA creates a directory for each image containing its diagnostics and results. The spot statistics are stored as a comma-separated text file, which can be easily imported into other tools for data analysis. An HTML-based interface allows the user to browse all results, and assists in their interpretation by prominently flagging results that indicate a problem with the image or its processing. This user interface enables the user to easily view diagnostics related to the entire set of images as well as each image individually.
|
CONCLUSIONS |
|---|
|
TOPAbstractINTRODUCTIONSOFTWARE IMPLEMENTATIONCONCLUSIONSREFERENCES |
|---|
The AMIA Toolbox for MATLAB provides a flexible, customizable tool for image analysis of microarray images. The modular, open-source nature of the software allows the user to understand how results are calculated and to add additional statistics and diagnostics specific to their particular application, a crucial capability that is not available in most microarray image analysis tools. This tool also provides the user with tools for microarray process quality control via extensive diagnostic statistics and images that pinpoint images and data of poor quality.
|
Acknowledgments |
|---|
This work was supported by the National Institutes of Health under Grant 5 R21 EB000980-02.
Conflict of Interest: none declared.
Received on May 12, 2005; revised on July 5, 2005; accepted on July 6, 2005
|
REFERENCES |
|---|
|
TOPAbstractINTRODUCTIONSOFTWARE IMPLEMENTATIONCONCLUSIONSREFERENCES |
|---|
Carlisle, A.J., et al. (2000) Development of a prostate cDNA microarray and statistical gene expression analysis package. Mol. Carcinogenesis, 28, 12–22[CrossRef][Web of Science][Medline].
Hojjatoleslami, S.A. and Kittler, J. (1998) Region growing: a new approach. IEEE Trans. Image Process., 7, 1079–1084[Medline].
Saeed, A.I., et al. (2003) TM4: a free, open-source system for microarray data management and analysis. Biotechniques, 34, 374–378[Web of Science][Medline].
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  H. Peng Bioimage informatics: a new area of engineering biology Bioinformatics, September 1, 2008; 24(17): 1827 - 1836. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. Novikov and E. Barillot Software package for automatic microarray image analysis (MAIA) Bioinformatics, March 1, 2007; 23(5): 639 - 640. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Doran, D. S. Raicu, J. D. Furst, R. Settimi, M. Schipma, and D. P. Chandler Oligonucleotide microarray identification of Bacillus anthracis strains using support vector machines Bioinformatics, February 15, 2007; 23(4): 487 - 492. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. P. Chandler, O. Alferov, B. Chernov, D. S. Daly, J. Golova, A. Perov, M. Protic, R. Robison, M. Schipma, A. White, et al. Diagnostic Oligonucleotide Microarray Fingerprinting of Bacillus Isolates J. Clin. Microbiol., January 1, 2006; 44(1): 244 - 250. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||