Microarrays

Microarrays are one of the workhorses of modern biology. Measuring transcript levels enables studies of differential expression - asking what the difference is, at the gene expression level, for example, between cancer tumor cells and normal cells.

Bruz Marzolf, who up 'til recently ran my local microarray facility, spoke recently, tracing the journey of microarrays through the full technology life-cycle, starting in 1995 with the publication of Quantitative Monitoring of Gene Expression Patterns with a Complementary DNA Microarray in Science. Bruz put microarrays in the category of a utility technology, but not quite to the point of commoditization as there remain major differences between manufacturers.

• Affymetrix, first to commercialize microarray technologies, is the 800 pound gorilla. Their photolithography process borrows from computer chip manufacturing and their standardized probe sets are well supported by tools such as Bioconductor. The technology is robust but producing the masks is quite expensive, thus custom arrays are not economical.
• Agilent, which spun out of HP, uses ink-jet technology. Custom arrays can be designed using Agilent's eArray software. Agilent arrays come in a variety of resolutions including 8x60k, 1x244k and 1x1m with 60mer probes.
• Illumina builds arrays out of beads coated in oligo probes. Beads are laid out randomly on the slides, necessitating a layout discovery step. These chips have extra redundancy to account for randomness in bead-probe count.
• Nimblegen's maskless photolithography process is more flexible for custom arrays. Nimblegen provides arrays in 385K, 4x72K, and 12x135K resolutions using 60mer probes. They emphasize high array-to-array data reproducibility.

As an aside, our group uses custom spotted arrays and Agilent arrays. We tried Nimblegen and found that inter-array consistency was excellent, but inter-probe consistency was not. Below we see the ribosomal RNAs and adjacent genes with total RNA measured by a custom Agilent array (in blue) plotted next to a custom Nimblegen array (in green). To be fair there might be other explanations for what we saw, but it certainly looks like there is significant variability between probes that we would expect to have identical readings.

In RNA-Seq: a revolutionary tool for transcriptomics (Nature Reviews Genetics, 2009), Zhong Wang, Mark Gerstein & Michael Snyder show this comparison between microarrays and RNA-Seq.

While RNA-seq, no doubt, has a higher dynamic range, does it really have less noise? Some folks say so. With tens or hundreds of thousands of probes, fairly dense coverage of whole microbial genomes is possible. If you know what you're looking for, microarrays are still cheaper. Discovery oriented work is going increasingly toward sequencing.

Links

• The technology life-cycle curve slide comes from a good talk by Simon Wardley given at OSCON in 2010.
• MicroArray Quality Control (MAQC) project, Shi et al, Nature Biotechnology, 2006
• RNA-seq: An assessment of technical reproducibility and comparison with gene expression arrays
• RNA-Seq: a revolutionary tool for transcriptomics
• Measuring differential gene expression by short read sequencing: quantitative comparison to 2-channel gene expression microarrays