DNA microarray: what it is, how it works and how to analyze the images

A DNA microarray is a glass slide, chip or membrane carrying thousands of spots, each holding a known DNA sequence (a probe). A sample’s DNA or RNA is labeled with a fluorescent dye and washed over the array; each labeled molecule sticks (hybridizes) to the spot with its complementary sequence. After washing, a scanner images the array, and the brightness of each spot shows how much of that sequence was in the sample. That is how one experiment measures the expression of thousands of genes at once.

This guide explains how a microarray works step by step, the main types, what the results can and cannot tell you, how a microarray compares with RNA-seq, and how the scanned image is turned into numbers.

A microarray experiment at a glance

The steps of a typical gene expression microarray experiment, what each one does, and what to check before moving on.

What a DNA microarray is

A DNA microarray, also called a DNA chip or gene chip, is a solid surface carrying an ordered grid of spots. Each spot holds many copies of one known DNA sequence, the probe. The sample being measured, the target, is labeled and hybridized to the array, and each probe captures its complementary sequence.

Two inventions made the method. In 1991, Stephen Fodor and colleagues showed that light-directed chemistry could build large, spatially addressable arrays of molecules on a surface (ref 1). In 1995, Mark Schena and colleagues printed complementary DNAs (cDNAs) on glass with a robot and measured the expression of 45 Arabidopsis genes at once by two-color fluorescence hybridization (ref 2). A year later, Lockhart and colleagues used high-density arrays of synthetic oligonucleotides to detect RNAs present at 1 in 300,000 and quantify them over more than three orders of magnitude (ref 3).

Array or assay? The array is the format: the grid of probes. The assay is the test you run on it: hybridizing a sample and measuring the signal. A “microarray assay” is an assay carried out on a microarray.

How a DNA microarray works, step by step

1. Design or choose the array. The probes decide what you can measure: one spot per gene, several per gene, or probes spread along the genome.
2. Extract and check the sample. For gene expression, extract RNA from each sample and check its quality and quantity before labeling.
3. Label the target. The RNA is usually copied into cDNA or cRNA and labeled with a fluorescent dye. On a two-color array, two samples are labeled with different dyes and mixed (ref 2); Cy3 (green) and Cy5 (red) are the usual pair. On a one-color array, each sample goes on its own array.
4. Hybridize. The labeled target is applied to the array and incubated so that each labeled molecule finds and binds its complementary probe.
5. Wash. Stringent washes remove target that is bound loosely or to the wrong probe.
6. Scan. A laser scanner excites the dyes and records the light from every spot at each wavelength, giving one image per channel.
7. Analyze the image. Software places a grid over the spots, finds each spot, measures its intensity, subtracts the local background and flags spots that cannot be measured.
8. Normalize and compare. The spot values are normalized to remove technical differences between dyes and arrays, then compared between samples with statistics (refs 4, 5).

Types of microarray

The main types of DNA and related microarrays, what is on the spots, what each measures and what each is used for.

Spotted cDNA arrays are usually read with two colors: each spot gives a ratio of one sample against the other (ref 2). In situ synthesized oligonucleotide arrays are usually read with one color: each array measures one sample, and samples are compared after normalization across arrays (refs 3, 6). Array comparative genomic hybridization (array CGH) uses the same two-color principle on genomic DNA: the ratio of test to reference DNA at each spot tracks copy number, which is how it maps deletions and amplifications (ref 7).

The same idea, a grid of capture spots read as an image, is used for proteins too: antibody and cytokine arrays are membrane or slide arrays read in the same way as a dot blot.

What a microarray result tells you

A microarray measures relative abundance. A spot’s brightness reflects how much labeled target bound to it, so the useful result is a comparison: gene A is higher in the treated sample than in the control, or region B has fewer copies in the tumor than in normal tissue. It does not give an absolute number of molecules without calibration.

From a gene expression experiment you can conclude which genes changed between conditions, by how much (usually reported as a log2 ratio or fold change), and which genes change together. Two caveats apply. First, a microarray only measures the sequences its probes were designed for, so it cannot find new transcripts. Second, probes can bind related sequences (cross-hybridization), and very weak and very strong signals fall outside the range a probe can measure (ref 8). Results for key genes are therefore often confirmed by another method.

Microarrays can be reproducible. In the MicroArray Quality Control (MAQC) project, coordinated by US FDA scientists, the same RNA samples run at multiple sites gave consistent results within each platform and largely agreed across platforms on which genes were differentially expressed (ref 9).

From image to numbers: microarray image analysis

Every microarray result starts as an image, and the image analysis decides what the numbers mean.

Gridding and spot finding. The software places a grid over the array layout and then finds the edge of each spot, so that every spot is measured over the area it actually covers.

Intensity and background. The signal of a spot is its intensity minus the background around it. Background comes from nonspecific binding and uneven fluorescence across the slide. Simply subtracting a local background estimate can produce negative values and noisy ratios for weak spots; model-based background correction performs better in comparisons (ref 10).

Quality flags. Saturated spots, dust, scratches, smears and spots that merge with their neighbors should be flagged and excluded rather than measured.

Normalization. Dyes label and fluoresce with different efficiencies, and arrays differ in overall brightness. Normalization removes these technical effects before samples are compared. On two-color arrays, a single global correction is not enough when the dye bias changes with spot intensity or position on the slide; intensity-dependent (loess) normalization addresses this (refs 4, 5). On one-color oligonucleotide arrays, normalization is done across arrays (ref 6).

Reporting. The MIAME guidelines (Minimum Information About a Microarray Experiment) set out what to report so others can interpret and verify the data, including the array design, the samples, the hybridization, the measurements and the normalization (ref 11).

Microarray vs RNA-seq

RNA-seq sequences the RNA in a sample instead of hybridizing it to probes. In a direct comparison on the same samples, the two methods gave highly correlated expression profiles, but RNA-seq detected more low-abundance transcripts, distinguished isoforms, identified genetic variants and had a broader dynamic range, because it does not depend on pre-designed probes (ref 8). The same study noted why microarrays remained in wide use: lower cost, simpler data storage and simpler analysis (ref 8).

Choose a microarray when the sequences you need are known, the sample numbers are large and the analysis needs to be simple and established. Choose RNA-seq when you need to discover new transcripts, isoforms or variants, or measure very low or very high expression.

Quantifying arrays with Phoretix Array

Phoretix Array is TotalLab’s software for quantifying array and plate images you capture yourself: DNA and RNA microarrays, cytokine array membranes, dot blots, slot blots, and ELISpot and FluoroSpot plates (ref 12). It handles the image analysis steps above in one place:

• Fit the grid. Place a measurement grid over the spots once, with built-in 96- and 384-well templates for plate formats (ref 12).
• Subtract background. Choose from several background subtraction methods and apply the same one to every spot (ref 12).
• Compare two channels. Open a two-color (Cy3/Cy5) scan as a multichannel TIFF and Phoretix Array works out the ratio between the channels for every spot.
• Use your controls. Define negative and positive control spots, normalize to one spot or to the average of several, and calibrate against standards of known amount (ref 12).
• Call presence. Set a threshold and mark every spot present or absent across the array (ref 12).
• Save it as a template. The next array with the same layout goes straight to results, and the template can be shared with colleagues and other sites (ref 12).
• Report. Export the results as CSV for statistics, or as a PDF report (ref 12).

Phoretix Array reads TIFF (including multichannel TIFFs, such as two-color Cy3/Cy5 scans), IMG, JPEG, PNG and BMP images from fluorescence, chemiluminescence and colorimetric detection, measures directly from the raw image, runs on Windows, macOS and Linux, and returns results in under a minute per image. With AuditSafe it adds audit trails, electronic signatures and image authenticity checks for 21 CFR Part 11 and EU Annex 11 (ref 12).

For whole-genome commercial chips, analysis normally runs in the platform’s own software and statistical packages, starting from the scanner’s output files (ref 6). Phoretix Array is for arrays where you have the image and need the spot values from it.

For membrane and antibody arrays read like dot blots, see the dot blot protocol.

If you quantified with Phoretix Array, the methods sentence reads: “Spot intensities were quantified using Phoretix Array (TotalLab Ltd, Newcastle upon Tyne, UK) after background subtraction and normalized to the [reference/positive control] spots.” Replace the bracketed part with the controls on your array (ref 12).

The video below runs the full Phoretix Array workflow on a spot array: fitting the grid, each background method, control spots, normalization, a calibration curve, presence and absence, and a saved template for the next image.

Get the numbers from your next array image

Open one of your own array images in a free Phoretix Array trial: fit the grid once, subtract background, normalize to your control spots and export the values.

Frequently asked questions

Q: What is a DNA microarray?

A: A glass slide, chip or membrane carrying thousands of spots of known DNA sequences (probes). A labeled sample is hybridized to it, each labeled molecule binds its complementary probe, and the brightness of each spot after scanning shows how much of that sequence was in the sample.

Q: How does a DNA microarray measure gene expression?

A: RNA from each sample is copied and labeled with a fluorescent dye, hybridized to probes for each gene, and scanned. The more RNA a gene produced, the brighter its spot. Comparing spot intensities between samples, after background subtraction and normalization, shows which genes changed.

Q: What is the difference between an array and an assay?

A: The array is the format: an ordered grid of probes or capture spots. The assay is the test run on it: applying a sample and measuring the signal. A microarray assay is an assay carried out on a microarray.

Q: What is the difference between one-color and two-color microarrays?

A: On a two-color array, two samples labeled with different dyes, often Cy3 and Cy5, are hybridized to the same array and each spot gives a ratio. On a one-color array, each sample goes on its own array and samples are compared after normalization across arrays.

Q: What conclusions can be made from a DNA microarray?

A: Which genes or regions differ between samples, in which direction and by how much, and which change together. A microarray gives relative, not absolute, amounts, only for the sequences its probes were designed to detect, so key results are often confirmed by another method.

Q: Is RNA-seq better than a microarray?

A: For discovery, often yes: RNA-seq detects more low-abundance transcripts, isoforms and variants and has a broader dynamic range. Microarrays remain useful where the target sequences are known and sample numbers are large, because they cost less and the data are simpler to store and analyze.

Q: How is microarray data normalized?

A: Normalization removes technical differences between dyes and between arrays before samples are compared. On two-color arrays, intensity-dependent methods such as loess correct dye bias that changes with spot brightness or position; on one-color arrays, normalization is done across all the arrays in the study.

References

1. Fodor SP, Read JL, Pirrung MC, Stryer L, Lu AT, Solas D. Light-directed, spatially addressable parallel chemical synthesis. Science. 1991;251(4995):767-73. doi:10.1126/science.1990438
   (Source for: light-directed synthesis of spatially addressable arrays, 1991.)
2. Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitoring of gene expression patterns with a complementary DNA microarray. Science. 1995;270(5235):467-70. doi:10.1126/science.270.5235.467
   (Source for: robotically printed cDNA arrays on glass; 45 Arabidopsis genes; two-color fluorescence hybridization.)
3. Lockhart DJ, Dong H, Byrne MC, et al. Expression monitoring by hybridization to high-density oligonucleotide arrays. Nat Biotechnol. 1996;14(13):1675-80. doi:10.1038/nbt1296-1675
   (Source for: in situ synthesized oligonucleotide arrays; detection at 1 in 300,000; quantitative over more than three orders of magnitude.)
4. Yang YH, Dudoit S, Luu P, et al. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variation. Nucleic Acids Res. 2002;30(4):e15. doi:10.1093/nar/30.4.e15
   (Source for: dye bias dependent on intensity and spatial location; global normalization not adequate; robust local regression (loess) normalization.)
5. Quackenbush J. Microarray data normalization and transformation. Nat Genet. 2002;32 Suppl:496-501. doi:10.1038/ng1032
   (Source for: normalization and transformation of microarray data before comparison.)
6. Irizarry RA, Hobbs B, Collin F, et al. Exploration, normalization, and summaries of high density oligonucleotide array probe level data. Biostatistics. 2003;4(2):249-64. doi:10.1093/biostatistics/4.2.249
   (Source for: normalizing one-color oligonucleotide arrays to one another; RMA.)
7. Pinkel D, Segraves R, Sudar D, et al. High resolution analysis of DNA copy number variation using comparative genomic hybridization to microarrays. Nat Genet. 1998;20(2):207-11. doi:10.1038/2524
   (Source for: array CGH; test to reference fluorescence ratio tracking copy number.)
8. Zhao S, Fung-Leung WP, Bittner A, Ngo K, Liu X. Comparison of RNA-Seq and microarray in transcriptome profiling of activated T cells. PLoS One. 2014;9(1):e78644. doi:10.1371/journal.pone.0078644
   (Source for: high correlation between platforms; RNA-seq better for low-abundance transcripts, isoforms and variants with broader dynamic range; microarray cross-hybridization and limited detection range; microarrays cheaper with simpler storage and analysis.)
9. MAQC Consortium; Shi L, Reid LH, Jones WD, et al. The MicroArray Quality Control (MAQC) project shows inter- and intraplatform reproducibility of gene expression measurements. Nat Biotechnol. 2006;24(9):1151-61. doi:10.1038/nbt1239
   (Source for: FDA-led project; intraplatform consistency across sites; interplatform concordance on differentially expressed genes.)
10. Ritchie ME, Silver J, Oshlack A, et al. A comparison of background correction methods for two-colour microarrays. Bioinformatics. 2007;23(20):2700-7. doi:10.1093/bioinformatics/btm412
    (Source for: local background subtraction causing negative intensities and noisy low-intensity ratios; model-based correction performing better.)
11. Brazma A, Hingamp P, Quackenbush J, et al. Minimum information about a microarray experiment (MIAME): toward standards for microarray data. Nat Genet. 2001;29(4):365-71. doi:10.1038/ng1201-365
    (Source for: the MIAME reporting standard.)
12. TotalLab. Phoretix Array product page. https://totallab.com/software/phoretix-array-array-assay-analysis-software/
    (Source for: applications including DNA or RNA microarrays; grid and 96 and 384 well templates; background subtraction methods; negative and positive controls; normalization to one or several wells; quantity calibration; presence assignment; templates shared between sites; CSV and PDF export; file formats; detection methods; raw image measurement; operating systems; under a minute per image; AuditSafe; how-to-report sentence.)