Microarrays present researchers with something of a catch-22: In order to find something, you have to know what you're looking for. If you're looking for new functional elements, your microarray needs to contain probes designed to target them. With many transcribed regions, protein-binding sites, and even genes still missing from genome annotations, however, even so-called "whole-genome" arrays with millions of features are likely to miss critical sequences.
But that is changing, thanks to rapid advances in the density of probes that can be packed into a single chip. In recent years, interest in hitherto-unexplored regions of DNA derided as "junk" has spurred the development of a new class of microarrays called "tiling arrays," in which probes are not designed to target known genes or promoters, but simply laid down at regular intervals along the length of the genome.
"It's much more unbiased. if you tile across a whole genome or...
ENCODE TAPS TILING TECHNOLOGY
Tiling arrays have been a key tool in the search for undiscovered functional elements, says Rick Myers, director of the Stanford Human Genome Center. "A huge amount of biology is dictated by proteins binding to a specific sequence of DNA. If you tile the entire genome, you can tell all the places on the genome where they are bound. It's a way in one experiment to find all the places, out of three billion base pairs, where you have a transcription factor bound."
In September 2003, the National Human Genome Research Project embarked on a mission to map of all of the functional elements in the human genome.6 The project, known as the Encyclopedia of DNA Elements (ENCODE), involves dozens of collaborating government and academic labs (including the Stanford Human Genome Center) as well as manufacturers. As a first step, the collaborators have been methodically probing a 30-megabase stretch of DNA, or 1% of the genome. Both Affymetrix and NimbleGen have developed tiling arrays specifically for the ENCODE project.
"It's taking one percent of the genome and hitting it hard with different technologies," says Peter Good, who codirects ENCODE. "We have a large amount of data coming out about transcription factor binding sites, histone modifications."
In his work for the ENCODE project, Iyer is using NimbleGen tiling arrays to corroborate the results of an experimental sequencing-based approach to identifying protein-binding sites on the genome. "Tiling arrays are a great way to validate what we are doing They're the best way for looking at transcription factor binding sites," he says.
Part of ENCODE's ongoing research involves detailed, multilab comparisons of tiling microarrays from different manufacturers, in hopes of quantifying the effect that factors like oligonucleotide length, tiling resolution, and platform have on various applications.
Both Affymetrix and NimbleGen use photolithography to create high-density oligonucleotide microarrays. But there are several important differences between the platforms.
With current-generation chips topping six million features, Affymetrix has far greater feature density than NimbleGen, whose chips contain up to about 390,000 features. The high feature-density allows Affymetrix to put an entire mammalian genome on seven arrays, compared to NimbleGen's 38.
However, NimbleGen's maskless technology allows it to be flexible in probe length within a single array, and to use longer probes. Affymetrix's arrays all have probes 25 bases long; NimbleGen's probes vary between 24 and 85 bases.
When it comes to oligonucleotides, "longer is better," says Stolc. "But you get diminishing returns after a certain point, because you lose sensitivity."7 NimbleGen's maskless photolithography technology allows for quick turnaround of novel array designs, enabling the company to offer custom arrays that tile at 10-base-pair resolution through a region of the customer's choice. NimbleGen will also make custom arrays for genomes that don't appear in the company catalog, making it a good choice for researchers who want to do microarray research on nonstandard organisms. NimbleGen's arrays cost roughly $1,000 per chip. For custom arrays, the company charges a design fee, usually between $500 and $3,000, depending on the complexity of the design. (Affymetrix declined to disclose pricing details for its GeneChip tiling arrays, though users have said they cost around $400 each.)
The resolution of the arrays – in other words, how many base pairs lie between the starting points of adjacent probes – varies widely, from a NimbleGen array that puts the entire human genome on a single chip with a probe every 6,000 base pairs, to an Affymetrix yeast genome array that tiles probes every five base pairs (all other Affymetrix tiling arrays have a 35 nucleotide resolution).
Affymetrix is now in the process of rolling out fully commercialized tiling arrays for human, mouse,
NimbleGen began offering a variety of tiling arrays earlier this year, including a whole-genome human CGH array with probes every 6,000 base pairs, whole-genome microbial tiling arrays, ChIP tiling arrays for human and mouse, and tiling arrays for
While the latest crop of high-density commercial tiling arrays are undoubtedly useful, says Myers, they're still not perfect. "The truth is that all of us would like yesterday an array that covers every base pair in the genome, that has high sensitivity, that never fails. None of them are even close to that," he says.
Nevertheless, tiling arrays have the capacity to take researchers further into the unexplored frontiers of the genome than did their predecessors, even the so-called "whole-genome" arrays, containing probes for all known genes from a particular organism, that were all the rage just a few years ago.
"Even a few years ago, like three or four, this totally seemed undoable, that you could actually look at the whole genome at such high resolution. [We thought it would] be years and years and cost a million dollars. It's actually happening. I think that's great," says Iyer.