Caitlin Smith has a B.A. in biology from Reed College, a Ph.D. in neuroscience from Yale University, and completed postdoctoral work at the Vollum Institute.
Researchers have used DNA microarrays to drive genetic analyses for more than two decades. Essentially grids of thousands or even millions of tiny spots of DNA printed onto glass slides, DNA microarrays enabled scientists to massively parallelize their research. No longer did they need to probe one gene or transcript at a time; with microarrays, they could extend their studies to the whole-genome or transcriptome level with little extra effort.
Naturally, the technology caught fire. And then along came next-generation DNA sequencing (NGS), a method that seemingly can do everything microarrays can, and more. In a sense, NGS changed the playing field by changing the types of questions researchers could ask. For instance, why look for phenotype-associated genetic markers when you can simply sequence the genome from end to end instead?
Still, rumors of the death of microarrays have been, as they say, greatly exaggerated, and many researchers still use them. Here’s why.
Cost discrepancies, and more
Many variables factor into cost estimates, including consumables, labor, time, sample type and (for NGS) the required read depth. But rough estimates put microarrays at about $100 (or less) per sample for basic genomic studies, and up to $300 per sample for more complex studies, such as splice variant analysis, says Kim Caple, senior vice president and general manager of clinical business at Affymetrix. NGS, in contrast, costs hundreds to thousands of dollars per sample—and that’s assuming existing access to an NGS platform—meaning that for analyses requiring many samples, such as clinical studies, NGS costs can quickly become prohibitive.
Microarrays also can save funds by targeting variants rather than the entire genome, “which is 99.9% identical in all individuals, [so it] adds cost but not information, because it is invariant,” says Mark Schena, president of Arrayit. A targeted DNA microarray runs from $10 to $100 per sample, says Schena, whereas “the whole human genome is typically $100 [to] $1,000 per sample.” (Targeting strategies are also available for NGS, to avoid the costs incurred by whole-genome sequencing, but they also add both cost and time.)
Microarrays also are faster to run than NGS experiments, and the equipment is broadly available. Another advantage: ease of data analysis. Free software can analyze microarray data quickly and reliably and “transform data into biological insights in minutes, compared to weeks or sometimes even months with [NGS],” says Anthony Schweitzer, head of bioinformatics in the expression business unit at Affymetrix. “This massively reduces the overall costs for each microarray project relative to [NGS].”
Recent microarray advances
Recent advances in microarray technology are challenging the notion that microarrays lag behind the newer NGS platforms. “There is a great deal of misperception about arrays vs. [NGS], especially with regard to what platform really needs to ‘catch up’,” says Caple. Microarray analysis can be performed with much smaller sample input (picograms of starting material) than NGS. Also, NGS requires sample-preparation steps that microarrays do not, which can lead to undesirable changes in sequencing coverage.
Just as in the sequencing space, microarray vendors also are advancing their technology. Here are some recent developments:
Genomics
Affymetrix has recently launched microarrays in a 384 well-format microarray to support larger sample genotyping studies that need higher throughput. Schena says Arrayit is “working on the $100 Genome” with its H25K whole-human-genome chip, which the company plans to launch this year.
Gene-expression profiling
Affymetrix’s new WT Pico Kit supports gene-expression profiling from as little as 100 picograms of total RNA or as few as 10 cells, says Dan St. Louis, general manager of the expression business unit at Affymetrix. “This is very enabling for researchers working with precious clinical samples or looking at single-cell transcriptomics.”
Copy number variation (CNV)
Affymetrix’s OncoScan whole-genome CNV assay gives a dynamic range of zero to 60 copies, using a starting sample as small as 80 ng of formalin-fixed paraffin-embedded tissue. Cancer researchers are using the OncoScan assay to study C-class tumors, which are driven by copy number variation, says Caple.
Single nucleotide polymorphism (SNP) detection
Microarray analysis is making strides in SNP detection. “With the advance of imputation techniques, arrays can be designed to provide tremendous genomic coverage,” says Mike Nemzek, vice president of strategic marketing in genotyping at Affymetrix. “An 800,000 SNP array can impute to many millions of SNPs.” SNP imputation is a method of statistically inferring alleles of hidden variants.
Diagnostics and clinical
Alessandro Borsatti, senior director of genomics product marketing at Agilent Technologies, says microarrays are being used for noninvasive clinical applications, such as circulating fetal cells in prenatal patients, or circulating tumor cells. Affymetrix’s CytoScan Dx Assay is designed for diagnosing postnatal developmental delay. Arrayit’s microarrays, constructed with ex situ oligonucleotide synthesis and contact printing, are being used to decode molecular bases of human diseases such as Parkinson’s disease, says Schena. Arrayit’s microarrays also are used for food-safety testing by the U.S. Department of Agriculture.
When to choose microarrays
So when is it better to choose microarrays instead of the newer NGS platforms? That depends on the application (and available instruments and resources). Microarrays are widely used in genomics, CNV, comparative genomic hybridization, methylation and gene-expression analyses. However, Scott Dewey, product manager in genetic sciences at Thermo Fisher Scientific, says microarrays for gene-expression studies “are experiencing rapid decline due the expansion of RNA-Seq and associated transcriptomics studies using NGS.”
Here are some common applications that play to microarrays’ strengths:
Gene-expression profiling
For gene-expression profiling, microarrays are faster, easier, cheaper and more familiar to some researchers. “NGS does have an advantage in dynamic range, although it is a matter of debate whether high-end dynamic range is all that important from a biological perspective,” says Nemzek. But NGS can lag behind in other respects (like sensitivity and specificity) for expression profiling—unless using moderate- or deep-read sequencing, which is far more expensive and slower than using microarrays.
Genotyping
For genotyping, you are better off using microarrays, especially on a large scale. “In human genotyping studies, where hundreds of thousands of samples are genotyped, microarrays are vastly more cost-effective, require less sample input and have a much higher throughput,” says Nemzek. Microarrays also are preferred in agrigenomics, where the faster time to result “is critical, due to timing of the planting season, in genetic screening for wheat, or [when] breeding salmon in aquaculture,” he says.
Genome-wide association studies (GWAS)
For GWAS, there are clear cost benefits to using microarrays. “Standard DNA-microarray GWAS prices are typically less than $1,000 for full-sample processing, arrays, consumables and labor,” says Dewey. “The cost for whole-genome sequencing will exceed this.”
CNV
According to Caple, NGS struggles with copy number analysis, regardless of cost. “DNA microarrays are still the gold standard for determining copy number changes,” says Borsatti. “Currently, NGS can reliably detect single nucleotide variants and small indels but cannot robustly detect larger copy number changes, such as single exon CNVs and large indels.” An exception, though, is long-read NGS technology, such as that developed by Pacific Biosciences.
Perhaps it is best not to view microarrays and NGS as opposing or competing methods, but as complementary, says Schena. “Microarrays are superior for genotyping and gene expression, because they offer greater dynamic range and sensitivity, better precision, better accuracy, much more rapid data analysis and are essentially immune to sample contaminants,” he says. Sequencing, on the other hand, “is important for identifying rare new variants in the genome and for constructing de novo sequences of plants and animals, which can then be used to construct DNA microarrays for affordable testing.”
Researchers today have the best of both worlds. After getting to know the strengths of both methods, they can benefit from the one that’s best suited to their work.
Image: iStockPhoto