The microarray image is your window into the biology you want to measure. The clearer that window is, the easier it will be to identify, interpret, and understand the phenomena you observe. Accurate imaging, including scanning and image processing, ensures that the data reported by the instrument provide the most accurate representation of the actual fluorescent signal on the array.

15.2 Scanning parameters Microarray scanner evolution

Microarray scanner technology evolved from scanning fluorescence microscopes that were used to image subcellular components. These microscopes combined high-powered excitation light sources, user-selectable emission filters, and a detection system to collect and store digital images of the fluorescent sample. Laser-based systems use motion control elements to scan the laser beam across the sample and photomultiplier tubes (PMTs) to collect emitted light one pixel at a time. White-light-based systems use mercury or xenon arc lamps to excite an entire field of view, and a charged-coupled device (CCD) to capture emitted light from the entire field of view. Both types of systems are used in microarray imagers today.

However, microarrays have different imaging requirements than cells and organelles. Microarray spots usually range from about 50 to 100 microns in diameter, which is extremely large compared to subcellular components. Unlike cellular imaging, the substructure of microarrays spots is of no interest (beyond optimizing spot uniformity). Microarray spots are deposited in known locations with plenty of separation between them, whereas cells and organelles are irregular in shape and location. Therefore, while cellular imaging requires ever-more powerful magnification and sub-micron pixel sizes to clearly resolve tiny structures, most

microarrays can be accurately imaged at 5- or even 10-micron pixel resolution. In addition, cellular microscopy has historically focused on the presence and location of specific elements of interest, rather than quanti-tation. (Note that this is changing, due in part to the development of flat-field imaging optics.) The primary purpose of microarrays is to quantify the signal from each spot. The data must be accurate and comparable over the entire 25 x 75-mm slide surface. Therefore, field uniformity is a critical parameter in microarray instrumentation.

How an image is acquired: scanner design considerations

Unlike light microscopy, which allows the viewer to look directly at the sample itself, fluorescence imaging requires a fluorescent label to be bound to the sample. When interpreting the results of microarray experiments, it's important to keep in mind that fluorescence imagers do not detect DNA, proteins, cells, or any other biological material - they only detect the fluorescent dyes that are bound to the biomolecules.

Fluorescence is the property of some molecules that absorb light of a given wavelength, and then emit light of a higher wavelength. Fluorescent dyes are characterized by their excitation and emission spectra. The excitation spectrum represents the efficiency with which the dye will absorb light over the given range of wavelengths. The emission spectrum indicates the probability that a photon of emitted light will be of a given wavelength. Thus, the peak of the excitation spectrum indicates the wavelength of light that is most efficiently absorbed by the dye, and the peak of the emission spectrum indicates the predominant wavelength of light that will be emitted as a result of the excitation. The difference between the excitation peak and the emission peak is called the Stokes' shift (1).

Not all fluorescent dyes are equally bright, even at their excitation and emission peaks. The brightness of dyes is determined by specific constants for each dye. The extinction coefficient indicates the efficiency with which the dye absorbs incident light (usually at the peak of the absorption spectrum). The quantum yield indicates the ease with which the dye molecule releases a photon of fluorescent light over the entire emission spectrum. The resulting fluorescence intensity is proportional to the product of these two constants. Therefore, equimolar amounts of two different dyes will not necessarily produce equally bright signals. Filter choices and excitation light properties (which also vary with wavelength) also contribute to differences in signal brightness.

In a fluorescence imaging system, excitation light is provided by either a halogen arc lamp or a laser. The light is delivered to the sample through a series of lenses and filters. The fluorophore on the sample emits light, which then travels through additional lenses and filters to the detector (a CCD or PMT). The analog signal from the detector is converted into a digital signal, which is then used to display an image of the sample on the computer screen.

Detailed descriptions and design considerations for white light and laser imaging systems have been discussed previously (2). This chapter focuses on critical instrument performance metrics, and some analysis methods that can be used to evaluate them.

Critical performance characteristics of a microarray fluorescence imaging system

Microarray scanners are complex instruments consisting of hundreds of individual parts, including light sources, detectors, circuit boards, moving parts, lenses, filters, wiring, and sensors. While the design and selection of each of these subunits is important, no single component defines the performance of the instrument. Ultimate instrument performance is determined by the integration of all of the subunits into a complete working system. The design of the electronic circuitry, alignment of optical elements, and behavior of moving parts all affect the data quality and long-term instrument performance. The specifications of individual components do not measure the ultimate function of combined subunits. The scanned image is the final output of the instrument; therefore the only way to characterize and compare the performance of the complete system is to compare scanned images. In addition, the appearance of the image on the screen is the result of the algorithms used to convert the analog fluorescent signal into a digital value, the color and display settings chosen, and even the monitor itself. Visual assessment of images can be misleading; therefore the signals must be quantified using appropriate background subtraction methods before valid comparisons can be drawn.

Scanner calibration

One slide does not constitute a microarray experiment. You may scan hundreds of slides over many months before reaching significant biological conclusions. Microarray scanners must perform consistently so that experimental results can be compared over time, and be validated and shared among different research groups. However, mechanical, optical, and electronic components have a finite lifetime, so instrument performance will change over time. Instrument calibration can ensure imaging reproducibility among multiple scanners of the same model over time.

Scanner calibration uses a known standard to set instrument output to predefined levels. The standard might be a precisely controlled light source or a fluorescent material. It must be stable such that it yields the same signal output after repeated long-time use. For example, GenePix® scanners are benchmarked at the factory to produce a specified signal output from a stable standard using defined scan settings. The test standards are a set of fluorescent materials that absorb and emit light consistent with each of the excitation and emission channels in the instrument. The standard is shipped with the instrument so that the user can invoke the calibration routine as often as they want to re-tune the instrument to the benchmark levels. Multiple instruments can be adjusted in the same way to produce the same benchmark signal levels.

Detection limit

A brighter image is not necessarily a better image. Absolute pixel intensity values will vary depending on different types and brands of detectors, variations in electronic signal processing, analog-to-digital conversion algo rithms, and other design differences. Color and display settings and even monitor settings can also influence the appearance of images scanned on different systems. Visual inspection is non-quantifiable and highly subjective, and is not a reliable method for detection limit comparisons on different instruments. Signal intensity and detection limit must be quantified using appropriate calculations and background subtraction methods (see below).

A detection limit performance specification should indicate the minimum signal that the system can quantify accurately. As a signal approaches the surrounding background level, the potential error in each measurement increases. In other words, as a spot fades into the background, so does your confidence in its existence. The signal-to-noise ratio (SNR) is the most reliable detection limit metric. The SNR calculation incorporates signal, the average background level, and the variation in background values to measure how clearly the signal can be distinguished from the background. For imaging applications, SNR is calculated as:

SNR = (Signal - Background)

(Standard Deviation of Background)

On most microarray scanners, spots may be visible below this limit; however, the accuracy of the measurement begins to diminish. As an analogy, consider looking for a 2-m-tall scarecrow in a cornfield. If all the cornstalks are 1 m tall, then the average background is uniform and lower than the signal (the scarecrow), so the scarecrow is clearly visible. If all the cornstalks are 2 m tall, then the signal is the same as the average background. Although the background noise is low, the average background is high so the scarecrow is not visible. Finally, if the cornstalks range in height equally distributed from 0.5 to 3 m tall, the average background is 1.75 m. The average background is lower than the signal, but the variation in height (i.e. the noise) will make it difficult to see the scarecrow. Thus, the signal, the average background, and the background variation must all be considered when determining detection limits in imaging applications. A commonly accepted criterion in many signal detection disciplines (including radio, electronic communications, trace chemical detection, and other fields) defines the minimum quantifiable signal at threefold greater than the background noise - that is the sample value for which SNR = 3 (3, 4).

Fluorescence imaging instruments do not detect DNA or proteins - they detect fluorescent dyes that are bound to the biomolecules. The detection limit of a fluorescence imaging instrument is measured in moles of fluo-rophore per square micron. The true detection limit can only be determined through careful quantitation of meticulously prepared dilutions of fluorescent dyes. In addition, the concentration of active fluorophore in a batch of dye varies according to batch preparation, age, and environmental conditions. Prior to arraying, the precise dye concentrations must be quantified on an independent platform such as a spectrofluorometer. The volume of solution that adheres to the slide during arraying must also be known. There can be no post-spotting washes or other treatment that may alter the amount of active fluorophore at each spot. The array must be used imme diately for detection limit determination because even the slightest decay in signal may cause the faintest spots to fade below the detection limit. These tests are time-consuming and are not practical for routine instrument comparisons.

An acceptable alternative to assess SNR differences among instruments is to compare any dilution series that covers a wide range of signal values. Several slide replicates should be scanned in alternating order on different instruments to assess and compensate for any photobleaching or slight differences among the replicates. The difference in SNR for identical spots near the background level gives a simple comparison of sensitivity across instruments, without knowledge of the absolute dye concentrations.

Field uniformity

Field uniformity is one of the most important specifications for microarray imaging. Microarray analysis entails measuring thousands of tiny signals on a relatively large field, with sufficient accuracy to allow spots to be compared among all locations on the array. A uniform imaging field ensures that the instrument is not contributing to regional variations that might bias the data and interfere with accurate comparisons.

The microarray substrate and surface matrix are the primary determinants of field uniformity. Most standard microscope slides are specified to about 40-|m flatness over the entire surface; that is, they may have hills and valleys as high as 40 | m. Optically flat slides for microarrays are also available. The slide surface variations can cause quantitative variations as the imaging plane comes in and out of focus. A scanner with a larger depth of field can better accommodate slide surface variations, ensuring accurate light collection over the entire scanned area.

Instrument components such as the slide holder, motion control mechanisms, the excitation source, and the illumination and collection optics can all affect field uniformity. In any microarray imaging system, all of these components must be precisely specified and aligned to ensure uniform illumination at all points on the sample surface.

A test standard to measure field uniformity must be more uniform than the instrument in question so that the sample itself doesn't contribute additional non-uniformity to the measurement. Such a standard doesn't yet exist in a microscope slide format. However, any fluorescent microarray can be used in a simple alternative test. You can scan the array in one orientation, rotate it 180┬░, and scan it again. A comparison of the signal intensities for identical spots in each scan quantifies field uniformity (Plate 6). Consistently lower signal in the second scan might indicate photo-bleaching. A third scan in either orientation can be used to quantify the photobleaching and subtract its contribution from the uniformity analysis. This rotation test is the most reliable measure of field uniformity using currently available tools. However, it is limited to variations that are asymmetrical with respect to the rotation. For example, a uniform hill or valley in the middle of the slide might go undetected.


A single microarray experiment is insufficient to reveal meaningful biological conclusions. Experimental replicates are critical to identify and eliminate experimental error and other variations, especially when an experiment uses many microarrays over the course of many months. Any complete microarray experiment should include array replicates, sample replicates, probe replicates (e.g. dye-swaps), and hybridization replicates. It is unwise to exclude replicates; however, you can reduce the cost of replicates and retain more data for analysis by minimizing the variation among them. If you devote some time to optimizing your protocols prior to starting a major experiment, and use care at each step in the microarray process, you will be rewarded with more reproducible results.

Instrument reproducibility is also an important parameter. Lasers and white-light sources need time to stabilize after igniting. In addition, extreme temperature and humidity fluctuations can cause variations in instrument behavior. To ensure repeatable results, users must observe the recommended warm-up times and operating conditions for their instruments. The best insurance against both short-term fluctuations and long-term signal drift is a calibration procedure such as that described above.

Signal repeatability of a microarray imaging system can be tested by repeatedly scanning a stable fluorescent standard (such as that described under "Scanner calibration" above) at identical scanner settings (Plate 7). Any reasonably stable fluorescent sample can also be used to test short-term scan-to-scan repeatability, although any photobleaching must be measured and subtracted.

0 0

Post a comment