State of the Microarray: Challenges and Concerns with Microarrays

Courtesy of CombiMatrix The research community's rapid acceptance of microarrays notwithstanding, technical challenges remain. Biochip developers continue to grapple with these issues while upgrading their offerings and adding new product lines in response to research trends. Perhaps the biggest challenge to microarray technology is standardization--ensuring that data collected from different microarray platforms can be accurately compared. Compounding this problem is the absence of a unified "language" for microarray analysis that would ease the exchange of microarray data between different groups. "There's very little consensus on how to analyze microarray data," says John Ambroziak of MWG Biotech in High Point, NC. Researchers are "trying to establish standards [to get] everybody speaking the same language." The Microarray Gene Expression Data Society (www.mged.org) has developed guidelines for the publication of DNA microarray data.1 These standards--collectively dubbed MIAME, for Minimal Information About a Microarray Experiment--are meant as a provisional solution to the broader problem of standardizing microarray data annotation and interpretation. Some journals--including Nature, Cell, and The Lancet--have adopted MIAME, thereby codifying what information authors must submit to have microarray data considered. Martin Potgeter, global product manager for microarrays for Netherlands-based QIAGEN, expects that this will lead more researchers to choose commercial arrays over the homegrown variety, and will inspire investigators to purchase complete array systems to ensure the reproducibility of their data. "[Journals] want to have proof that signals are not false positives," says Potgeter. "The only chance to get high-quality data is to do [experiments] in replicates or ... by getting a high-quality chip from a good supplier." Stanford University biochemist Patrick Brown agrees that data quality and reproducibility are critical but says that commercial microarrays have not yet made these better, primarily because of their prohibitive cost, "which induces experimenters to cut corners on controls and repetitions of experiments." Frederick, Md.-based SuperArray is developing an alternate approach, reporting gene expression levels not with arbitrary numbers, but in terms of copies per microgram of total RNA or per cell. According to SuperArray senior scientist Xiao Zeng, such technology would allow data generated from SuperArray's products by different labs using different samples to be directly compared. PROBE PARTICULARS In 2001 scientists at Rosetta Inpharmatics and Agilent Technologies investigated the effects of probe length and other parameters on the specificity of oligo microarrays generated by an inkjet printer. They found that longer, 60-mer probes reliably detected low-abundance transcripts and in some cases performed better than single, shorter probes, as short oligonucleotides lacking a spacer to move the probe into the more favorable solution state tend to suffer from inefficient hybridization.2 Companies that produce arrays with longer probes view this study as a validation of the approach. "I think the market is becoming more confident in long oligo arrays," says MWG Biotech's Ambroziak. "You have a hard time getting true gene specificity with shorter oligos." MWG manufactures arrays of purified 50-mers, and according to Ambroziak, its competitors in the premade array field are generally adopting this strategy, too. One such company is Palo Alto, Calif.-based Agilent, which recently launched a line of catalog oligo microarrays printed with 60-mer probes. Other companies, including BD Biosciences Clontech of Palo Alto, Calif., and QIAGEN, sell arrays spotted with 70-mer probes. But longer oligos are more costly to make and more difficult to accurately synthesize. Amersham Biosciences in Piscataway, NJ, offers a new type of array that reportedly eliminates hybridization problems by putting the oligo probes into a three-dimensional gel-like matrix that mimics a solution-phase environment. The CodeLink™ system,3 which Amersham recently acquired from Motorola Life Sciences, represents each gene with a functionally validated, 30-mer sequence. Chockalingam Palaniappan, director of research and development, gene expression, explains that since every base in this environment participates in hybridization, the arrays offer higher reproducibility, specificity, and sensitivity than standard glass slide arrays. Another approach to microarray manufacturing is that pioneered by Affymetrix of Santa Clara, Calif., which synthesizes short oligonucleotide probes 20-25 bases long directly on the surface of chips using a combination of photolithography and combinatorial chemistry. Affymetrix uses a set of probe pairs to represent each transcript, including multiple perfect match probes and corresponding mismatch probes. According to Elizabeth Kerr, senior director of marketing, gene expression, this approach provides higher sensitivity and specificity than using a single longer oligo, reducing background noise and minimizing false-positive results. While Affymetrix's spot density is already high--up to 500,000 features per chip--there is still room for improvement. Greg Yap, Affymetrix's senior director of marketing, explains that decreasing feature size to increase chip density has been and remains a corporate priority, and is possible because Affymetrix's manufacturing method exploits technology used by the semiconductor industry, which currently synthesizes features that are orders of magnitude smaller. With smaller features, says Yap, more information could be packed into a given amount of space, driving down the cost of the arrays. Affymetrix plans to release several new high-density array products this year, including CustomSeq, for large-scale resequencing projects, and Mapping 10K, the first commercially available array that allows for genotyping 10,000 single nucleotide polymorphisms. Courtesy of PHASE-1 Molecular Toxicology BE FLEXIBLE Microarray developers note that array flexibility, or the ability to use the arrays for high- or low-content applications, is also a priority for cost-conscious customers. Researchers who wish to concentrate on one specific biological pathway or study a limited number of genes can choose products from companies such as SuperArray and OriGene, which specialize in low-density, focused arrays. Another approach is to increase the chip's feature density and break it into smaller or larger pieces, depending on the application; Affymetrix favors this approach. An alternate slant is to produce modular chips, as QIAGEN has done with its new SensiChip arrays.4 Each SensiChip Bar contains six arrays separated by micro-fluidic hybridization chambers. Users can choose to run six different experiments on a single bar for low-density applications, or they can increase the density of the bar by dividing the content on several arrays, allowing one, two, or three experiments per bar. Potgeter explains that customers who are interested in studying low numbers of genes will look towards customized low-density arrays to minimize costs and analysis time. "The lower the density of the SensiChip, the more cost-effective it is for the customer, because the customer can divide the price per bar by a factor of up to six," he says. PROBING THE 'OMES Microarrays help researchers quantitate gene expression by measuring the relative level of mRNA expression, but alternative splicing of mRNAs makes mRNA expression patterns more complicated than gene expression alone. "Thinking in terms of 'gene expression' is often inappropriate, since microarrays are measuring the expression of mRNAs," says Alon Amit of Tel Aviv-based Compugen. Therefore, microarray probes should be designed with the transcriptome in mind. "It could be the case that differences in expression ... at different times or in different cells ... are due to differential expression of specific splice variants of the given gene, not necessarily overall expression of the gene, and that influences both the design [of the probes] and the interpretation of the results," says Amit. The choice of probe sequences has been driven by expressed sequence tag data obtained through public databases such as UniGene and GenBank, but Amit says polymorphisms, chimeric sequences, and the presence of antisense genes can limit the accuracy of such data. As a result, microarray manufacturers are forming partnerships with bioinformatics companies to develop probes that more accurately reflect mRNA variation. For example, Compugen has collaborated with Sigma-Genosys and has microarray-design agreements with Pfizer and Novartis. Microarrays themselves can be used to identify splice variants, as well as small noncoding RNAs that play a role in gene regulation.5 Recognizing that mRNA expression levels do not always accurately reflect the expression of corresponding proteins, researchers are increasingly combining gene expression and proteomics data. Scientists at the Institute for Systems Biology in Seattle, for instance, recently used an integrated genomics-proteomics approach to analyze the yeast galactose utilization pathway.6 The development of technologies to increase the throughput of protein interaction studies would clearly aid similar approaches to the study of biological pathways. As a result, companies currently offering DNA microarray technologies have begun migrating toward the high-throughput proteomics market. For example, Boston-based PerkinElmer, known for its MICROMAX™ line of cDNA arrays, now offers a complete proteomics package, including slides, printing and scanning equipment, and labeling and amplification reagents for the production of protein microarrays. Likewise, BD Biosciences Clontech, which manufactures Atlas™ gene expression microarrays, now supplies antibody microarrays. Other companies are joining the protein array bandwagon. Sunnyvale, Calif.-based TeleChem International's ArrayIt division is currently developing protein microarray technology and plans to release protein array accessories designed to be compatible with the company's SpotBot arrayer. "The direction people are moving towards is ... microarray-based proteomics," says Robin Stears, TeleChem's director of microarray technology. "There needs to be some development on the end of detection systems and slide formats and chemistries that are compatible with protein systems." To this end, TeleChem is developing scanners for protein microarrays. Aileen Constans (aconstans@the-scientist.com) is a contributing editor. References 1. C.A. Ball et al., "Standards for microarray data," Science, 298:539, Oct. 18, 2002. 2. T.R. Hughes et al., "Expression profiling using microarrays fabricated by an ink-jet oligonucleotide synthesizer," Nat Biotechnol, 19:342-7, 2001. 3. J.M. Perkel, "CodeLink enters the fray," The Scientist, 16[12]:47, June 10, 2002. 4. H. Sussman, "QIAGEN reveals a sensitive side," The Scientist, 16[13]:37, June 24, 2002. 5. D.D. Shoemaker, P.S. Linsley, "Recent developments in DNA microarrays," Curr Opin Microbiol, 5:334-7, 2002. 6. T. Ideker et al., "Integrated genomic and proteomic analyses of a systematically perturbed metabolic network," Science, 292:929-34, 2001.

State of the Microarray: Challenges and Concerns with Microarrays

Interested in reading more?

Become a Member of