Roche Applied Science announced last week it has begun selling 454 Life Sciences' Genome Sequencer 20 the first high-speed, non-Sanger sequencing system to be made commercially available. It may be only a matter of time before next-generation sequencing muscles in on SNP-genotyping in some research and diagnostic applications, but industry observers say that cost and data quality will ultimately be the driving factors in widespread adoption of the new technology.
"Genotyping cost is always ultimately going to be limited in how far down you can go [in cost], where the sequencing technologies might not, because they might be relying on different properties of DNA," Chad Nusbaum, co-director of the Genome Sequencing and Analysis program at the Massachusetts Institute of Technology's Broad Institute, told Pharmacogenomics Reporter.
It remains unclear when sequencing will begin to displace existing SNP genotyping from some applications, and to be clear, Roche and 454 are not currently marketing the sequencing technology as a genotyping competitor. But should recipients of the US National Human Genome Research Institute's "$100,000 genome" and "$1,000 genome" rants succeed in reaching their five- and 10-year goals, respectively, the two may come into conflict in the near future. It won't be a complete revolution over SNP genotyping, however not for a while, at least. In well-understood genetic polymorphisms, there may not be a substitute for SNP tests for years.
"If you can be two-fold cheaper than an existing technology, then I think that's when people are going to start to really jump on it."
Defining the territory over which genotyping and sequencing tools will compete can be a slippery job. "There are many different ways to frame the question this is why if you go to one of these new sequencing-technology companies, they're going to tell you, 'Yes, it's dramatically cheaper' to use their technology than anything you ever do because the calculations you can make are so flexible," said Nusbaum. Sequencing can sound more attractive than SNP genotyping depending on how one approaches problems such as the cost of PCR primers, or whether a single read will reliably identify alleles, he said.
"The hope is that by the time we get to practical care, there will be a limited subset [of physiologically relevant nucleotide bases], but while we're in studies it's very hard for anyone to point to a base pair and say, 'That base pair is not important for anything,'" said George Church, a professor of genetics at Harvard Medical School and director of the Center for Computational Genetics.
On the Costs
"Data quality per cost" is the key factor over which sequencing and genotyping compete, said Jeff Schloss, program director of technology development at the US National Human Genome Research Institute. The going rate for a mammalian genome-sequencing job of approximately 3 billion bases is about $10 million, but that figure also includes the cost of assembling short stretches of sequence, he said.
Because the 454 sequencing technology is not yet mature, a sample head-to-head comparison against Illumina's cheapest genotyping technology in identifying alleles might yield a price advantage for genotyping of "a factor of two to three," as long as primer costs are left out of the equation, said Nusbaum. "We're hoping that the cost of running [454's] machine is going to come down ten-fold in the next year or two that's going to make it very competitive in a number of areas, not just genotyping."
Neither Roche nor 454 returned calls for this article.
One of genotyping's advantages is the simplicity of its data a sequencing run will produce an enormous amount of data to pore over, David Ginzinger, Applied Biosystems' director of scientific operations, told Pharmacogenomics Reporter. Although ABI has recently announced its intention to develop a next-generation sequencing technology of its own, he expects capillary electrophoresis-based Sanger sequencing to remain standard due to its lower error rate and high market penetrance.
In addition to avoiding a data deluge, genotyping's advantages include its high efficiency using tag SNPs, its diminishing cost, and its high data quality, said Bill Craumer, an Illumina spokesperson. Sequencing the whole genome is indeed interesting, but your error rates have to be much lower than genotyping given the huge number of data points you're looking at. "So there may be fine mapping applications where the base-to-base detail of sequencing would provide better information, especially if you had identified a region of interest and wanted to explore," he said.
"There's a price point for everything that's really what's going to drive the decision in the end," Nusbaum said. Supposing no other advantages, "if you can be two-fold cheaper than an existing technology, then I think that's when people are going to start to really jump on it" for large-scale applications. "Once you start to be five- to 10-fold cheaper, even if your quality is lower, then people sit up and take notice," as the public did with digital photography, he said.
The break-even point, where SNP genotyping and sequencing cost the same, assuming similar data quality, is roughly where the cost of one SNP equals the cost of sequencing the average amount of meaningful DNA represented by a single SNP. "Let's say the whole gene you're interested in has one SNP, but has 10,000 bases that is about break-even," said Church.
The method developed by Solexa costs approximately $100,000 per three billion bases, but that estimate applies only to resequencing, according to a company spokesperson.
A spokesperson for Helicos said that the company would not disclose estimates of the cost of sequencing using its system until the platform reaches the market.
Estimates of the cost of sequencing using new technologies may not be completely meaningful, Nusbaum cautioned. But it really doesn't matter very much, he said. "I don't care who's the cheapest, as long as they're in competition with each other and it drives the cost down."
As sequencing technology stands now, 10 or 20 reads are required to attain data as reliable as that produced by genotyping, said Nusbaum. "That's why the cost has to be so much lower per base [in sequencing]" to get information comparable to a SNP reading using genotyping, he said. "If somebody came out with a new technology next year that was a 100-fold cheaper, we could do 20-fold more oversampling, greatly increase our accuracy, and lower our cost five times."
The standard error rate with Sanger sequencing is approximately one error for every 10,000 bases read, said Church. "Some of the new methods are very noisy, and so you have to go to very high coverage, and you still get 10-3 to 10-4," he said. A method published by Church and colleagues requires three-fold oversampling to achieve an error rate of about 10-6 to 10-7, he added.
The current cost of genotyping is about five cents per SNP using Sequenom's platform, said Harry Stylli, the company's CEO, in a recent conference call with investors. The company plans to drive the cost per genotype down to three cents as soon as the third quarter of 2006, and eventually reach the elusive price of one cent per SNP, he said.
Home-Court Technological Advantages
A drawback inherent in SNP genotyping is that it depends on some knowledge of gene-pool variation for each gene of interest. "To the extent that the variation in any sample conforms to that model, you have a pretty good chance of identifying it" with SNP genotyping, said Schloss. But studies become difficult when there is a lot of sample deviation from the known range of genetic variation, he said. "If you don't know what your SNPs are, then you have to go to sequencing, and I think that at least in the short term is where the [genotyping and sequencing] technologies have their strengths," Nusbaum said.
For disease areas in which there are a number of disorders that are difficult to distinguish from one another by symptoms, such as motor-neuron disorders, cheap sequencing would prove useful in diagnosis through polymorphism discovery in dozens or hundreds of genes, Nusbaum said. Asthma, diabetes, and hypertension also seem attractive areas in which to use to his approach, he said.
Cancer is a good bet for sequencing's diagnostics beachhead. Tumor profiling is "not something that you can't do with genotyping at all," while the new technologies, such as those of 454, Solexa, and Helicos, can provide proportional representations of the different genes and alleles present in a heterogeneous tumor-cell sample, said Nusbaum.
Chris Womack ([email protected])