A particular tool that has been shown to be very useful in genomics is the DNA microarray. In its most general form, the DNA array is a substrate (nylon membrane, glass, or plastic) on which DNA is deposited in localized regions arranged in a regu- lar, grid-like pattern. Two main approaches are used for microarray fabrication:

further divided into photolithography, ink jet printing, and electrochemical synthesis. The photolithographic approach (Affymetrix) is similar to the VLSI fabrication pro- cess: photolithographic masks are used for each base. If a region should have a given base, the corresponding mask will have a hole allowing the base to be deposited at that location. Subsequent masks will construct the sequences base by base. This tech- nology allows the fabrication of very high-density arrays but the length of the DNA sequences constructed is limited1. The ink jet technology (e.g., Agilent and Protogene) is similar to the technology used in ink jet color printers. Four cartridges are loaded with the A, C, G, and T nucleotides. As the print head moves across the array substrate, specific nucleotides are deposited where they are needed. The elec- trochemical synthesis (CombiMatrix) uses small electrodes embedded into the sub- strate to manage individual reaction sites. Solutions containing specific bases are washed over the surface and the electrodes are activated in the necessary positions. In the deposition-based fabrication (e.g., Clontech and Corning), the DNA is prepared away from the chip. Robots dip thin pins in the solutions containing the desired DNA and then touch the pins onto the surface of the arrays. Small quantities of DNA are deposited as spots on the array. Unlike in situ manufacturing, spotted arrays can use small sequences, whole genes, or polymerase chain reaction (PCR) products.

The DNA array is subsequently probed with complementary DNA (cDNA) obtained by reverse transcription of mRNA extracted from a tissue sample. This DNA is fluorescently labeled with a dye and subsequent illumination with an appropriate

In order to limit the overall probability of an error, one needs to limit the length of the sequences. To compensate, many short sequences from the same gene can be synthesized on a given array. The particular sequences must be carefully chosen to avoid cross-hybridization.

light source will provide an image of the array. The intensity of each spot or the aver- age difference between matches and mismatches can be related to the amount of mRNA present in the tissue which in turn is usually correlated with the amount of protein produced by the gene corresponding to the given region. The labeling can also be done with a radioactive substance. Figure 1 illustrates the microarray process. In a multichannel experiment, several probes are labeled with different dyes and used at the same time in a competitive hybridization process. A large number of expression values are obtained after processing the image(s) of the hybridized array. Typically, one DNA chip will provide expression values for thousands of genes.

Although microarrays have been used successfully in a range of applications including sequencing and single nucleotide polymorphism (SNP) detection, most applications are related to gene expression. Typical examples include comparing healthy and malignant tissue, studying cell phenomena over time or studying the effect of various factors on the global pattern of gene expression.

Due to their nature, microarrays tend to be very noisy. Even if an experiment is performed twice with exactly the same materials and under exactly the same condi- tions, it is likely that after the scanning and image-processing steps, many of the same genes will probably be characterized by different quantification values. Noise is introduced at each step of various procedures2: mRNA preparation (tissues, kits,