High Resolution Imaging for Accurate Microarray Quantitation
Microarray elements are formed by small droplets of liquid drying on a solid substrate such as a micro-scope slide. The array elements or spots typically range in size from 25 µm to 500 µm in diameter, with most microarray spots measuring approximate-ly 100 µm. Accurate analysis of 100 µm spots requires detection systems such that each microar-ray spot be imaged into as many significant pixels as possible. Ahigher number of significant pixels per spot can correct for edge artifacts and other non-uniformities (i.e. dust, etc.) in the quantification of the fluorescence signal. As a rule of thumb, the pixel dimension or spatial resolution of a microar- ray scanner should be no larger than approximately 1/8 to 1/1 0 of the diameter of the smallest microarray spot (i.e. 10 µm pixel resolution for 100 µm microarray spots).
Two methods to scan and quantitate fluorescent microarrays are CCD-based imaging and confocal laser scanning systems. Multi-element detector arrays, such as a CCD cameras, are configured to stare at a small area that has been flood-illuminated, and to provide an image divided into pixels directly by the CCD detector. The limitation to this approach is generally lower pixel resolution (i.e. approximately 30 µm to 50 µm) and therefore, lower image quality and fewer statisti-cally significant pixels available for accurate quan-titation. 50 µm resolution provides approximately 2 pixels in the diameter of a 100 µm spot and, in the best case, from 0 - 1 signif-icant pixels. 30 µm pixel res- olution provides 3 pixels in the diameter of a 100 µm spot and only 5 significant pixels.
In contrast to CCD imaging systems, confocal laser scanners provide high resolution imaging by focus-ing the excitation beam to a point about the size of a pixel (e.g. 25-100 µm 2 ), and collecting emissions from just that small area with a single-element d e t e c t o r. Commercially available confocal imaging systems such as the ScanArray 3000 provide 10 µm pixel resolution, and next generation ScanArray systems will provide 5 µm pixel resolution. This resolution range maximizes the number of significant pixels and optimizes image quality and quantitive analysis. 10 µm resolution provides 10 pixels in the diameter of a 100 µm spot and approximately 70 significant pixels. 5 µm resolution pro-vides 20 pixels in the diameter of a 100 µm spot and a total of approximately 320 significant pixels.
Furthermore, a large number of significant pixels can also aid in the quantitation of spots that have a high degree of non-uniformity, dust, or other spot-ting artifacts, i.e. doughnut effect as the "defective" pixels can be ignored by the quantitation algorithm, while retaining enough "good" pixels for accuracy. Fin ally, a greater number of statistically significant pixels are particularly important when using signal histogram quantitation methods (robust algorithms adopted by Stanford University and found in sever-al commercial software packages) for microarray analysis. The histogram quantitation method incor-porates user-defined thresholds on background and signal histograms to define mean signal and back-ground values. The pixels that contribute to more accurate data are ones that consist entirely of spot signal data (statistically significant) and not inter-mediate value pixels consisting of signal and back-ground data.
Provided by: Packard BioChip Technologies