By Julia Karow
As part of the first detailed description of its semiconductor-based non-optical sequencing technology, Life Technologies' Ion Torrent last week laid out a path to a sequencing chip with more than a billion sensors, almost 200 times more than the current 316 chip.
As proof of concept for human genome sequencing, the company also published the draft genome of Intel co-founder Gordon Moore, which required about 1,000 runs on its Personal Genome Machine with the 314 chip. The paper, which also described the sequencing of three bacterial genomes, was published by Ion Torrent founder Jonathan Rothberg and colleagues in Nature last week.
A company representative stressed that the PGM is not yet ready for human whole-genome sequencing, while opinions among researchers differ on how useful the platform might be for that application in the future.
Data for the paper was generated at the beginning of the year using the 314 chip, which has about 1.2 million wells, or fluid-accessible sensors. Since then, the company has launched the 316 chip, which has 6.1 million wells, and plans to release by the end of the year the 318 chip, which will have 11 million wells.
The paper's data might be outdated by now, but it helps drive home the company's point that the technology is rapidly improving in throughput. In addition, it contains "forward-looking research experiments that show that it can scale quite a bit further," according to Maneesh Jain, Ion Torrent's vice president of marketing and business development.
Scaling Semiconductor Sequencing
Instead of sensing photons, like existing sequencing technologies, the Ion Torrent technology detects protons that are released when a DNA polymerase adds a natural nucleotide during DNA synthesis.
The company uses an ion-sensitive field-effect transistor as its sensor since it is "most applicable to our chemistry and scaling requirements because of its sensitivity to hydrogen ions, and its compatibility with [complementary metal-oxide semiconductor] processes," according to the authors.
The paper acknowledges that other researchers have attempted to sequence DNA electronically in the past, but "none of [these attempts] produced de novo DNA sequence, addressed the issue of delivering template DNA to the sensors, or scaled to large arrays."
Ion Torrent's current chips — the 314 and the 316 — contain wells with a diameter of 3.5 micrometers, their centers spaced 5.1 micrometers apart, that sit on top of a proton-sensitive tantalum oxide layer. Underneath is an array of sensor elements, each connected to an underlying ISFET via a floating metal gate. The chips are packaged with a disposable polycarbonate flow cell that isolates the sample and fluids from the electronics.
Both the 314 chip and the 316 chip use three transistors per sensor, but the 316 chip has a larger area. The 318 chip will only have two transistors per sensor, which will be packed more densely.
In order to prove that they can pack even more wells on a chip, Ion Torrent scientists shrunk the wells to 1.3 micrometers, their centers spaced just 1.68 micrometers apart. In the paper, they show that they can use these wells to generate sequence data for the first four bases of a template, using 1-micrometer template beads. These smaller wells, combined with a larger chip area, will allow the company to manufacture chips with 165 million wells, using two transistors per sensor.
[ pagebreak ]
The scientists speculate in the article that by increasing the chip area further, shrinking the wells to 0.5 micrometers, and using only one transistor per sensor, they might be able to produce chips with 1.1 billion sensors in the future, about 180 times more than the 316 chip and about 90 times more than the 318 chip.
However, Ion Torrent is currently not saying when or how it plans to commercialize these new chip designs.
Ion Torrent scientists also describe in the paper how they used about 1,000 of the314 chips to sequence the genome of Gordon Moore, author of Moore's Law (IS 3/1/2011), to a mean coverage of 10.6-fold. They called about 2.6 million SNPs in his genome and found that he has about 3,400 of the 7,700 deletions and inversions discovered by the 1000 Genomes Consortium.
Based on the current $99 list price for the 314 chip, Moore's low-coverage genome sequence would cost almost $100,000 in chips alone today, not counting reagents, while companies like Illumina and Complete Genomics are offering high-coverage human genomes for about $5,000, when ordered in bulk.
Jain noted that later this year, researchers will be able to generate the same amount of sequence with 30 of the company's 318 chips, which will have a list price of $500, reducing the chip cost for a low-coverage human genome to $15,000.
According to Jain, the project was not intended to suggest that the PGM is ready for human genome sequencing today. "Nobody is proposing that you should use the PGM and 314 chips to do the human genome," he said. "That was not the point of the paper, and somebody saying that it's too expensive, too cumbersome, or not enough coverage is missing the point."
Rather, he said, the goal of this project was to show that in principle, the Ion Torrent technology can be used to sequence a human genome "with all its complexity." In addition, he said the company wanted to honor Gordon Moore, and his role in semiconductor development, by making him the first human to be sequenced using semiconductor-based sequencing technology. "So yes, it was a little tour de force, but … it was important to sequence his genome."
Scientists differ in their assessment of how useful the PGM might be for sequencing human genomes or other large and complex genomes in the future.
According to Stephan Schuster, a professor of biochemistry and molecular biology at Pennsylvania State University, the advent of the 318 chip, with has an expected output of a gigabase, will make the platform applicable to sequencing mammalian and large plant genomes de novo, in combination with other sequencing technologies.
Using its two PGM instruments, his lab could "easily" generate eight-fold coverage of a human genome in a four-day week, assuming three runs per day, he said. This is the same amount of sequence they generated two years ago for an African genome on the 454 platform, he noted, which took close to two months and four instruments (IS 2/23/2010).
[ pagebreak ]
Schuster, whose lab is an early-access customer of the Ion Torrent technology, said that the PGM provides "remarkably even" coverage across the genome. Furthermore, the majority of errors are indels, and the substitution error rate is low, making it an ideal complement for de novo genome sequencing on platforms with different error profiles. "You do a combination of, maybe, a MiSeq or a lane of HiSeq and the Ion Torrent, and you will get the best of both worlds," he said. "This will allow you to assemble relatively fast and low-cost, very workable draft versions of genomes."
Others are trying to achieve the same with Illumina data alone, he said, "but they will always have the limitations that come with the use of a single sequencing platform."
For Matthew Bainbridge, a researcher at the Human Genome Sequencing Center at Baylor College of Medicine, the PGM is interesting for human genome sequencing because of its speed and because it has the potential to generate long reads, maybe as long as Sanger reads. "Once you see this technology start scaling up to 500 base reads, which will be significantly longer than Illumina reads, then I think you will see a lot of interest in this technology," he said. Longer reads would allow the platform to call long insertions and deletions, for example, which he said are difficult to detect with Illumina's technology.
The high speed of the PGM, and the ability to run several of the relatively inexpensive machines in parallel, might soon enable researchers to complete a human genome in a day, which could be important for clinical applications, "where you had kids coming in sick and you needed their genome yesterday," he said.
"Even if they end up being a little bit more expensive, if they can call mutation types that you can't call using HiSeq, and they are a lot faster, then there is going to be a market there for them," Bainbridge said.
However, he pointed out that the platform — like 454's — tends to overcall indels near homopolymer runs. "That's going to be, I think, the real challenge for Ion Torrent's technology, to overcome their indel problem," he said.
Regarding the Moore genome, Bainbridge said it appears to miss a lot of heterozygous single-nucleotide variants because it was not sequenced deeply enough. However, the company was able to validate most of the SNPs it found, he said. "They showed that they can do quality calls off this technology."
Not everyone agrees about the technology's promise for human-genome sequencing, though. "The Ion Torrent technology has a long way to go before it's going to be useful for human sequencing, and my bet would be that a newer technology will be available before it gets up to speed for human sequencing (if it ever does)," said the director of a DNA sequencing facility that focuses on human whole-genome sequencing in an e-mail message, who asked to remain anonymous.
He added that he doubts Ion Torrent will achieve long enough reads with an acceptable error rate to help with de novo alignment, and that Ion Torrent's limitation will likely be its chemistry, not its ability to scale its chips.
For now, the PGM will continue to target applications that require between 10 megabases and 1 gigabase of data output, Ion Torrent's Jain said, such targeted resequencing, amplicon sequencing, and RNA-seq. Whether Ion Torrent will develop commercial chips that will enable the instrument to sequence human genomes is unclear at this point. "What we have shown is that the technology has no fundamental limitations that we know of," Jain said.
"From a scientific standpoint, it's a seminal paper which shows which direction things can go," he said. "How you translate that into products, and which products, is complicated and too early to speculate."
Have topics you'd like to see covered in In Sequence? Contact the editor at jkarow [at] genomeweb [.] com.