Researchers in George Church’s laboratory at Harvard University are developing a second-generation high-throughput sequencer that they say will cost around $100,000 and be robust and easy to use.
Based on the prototype polony sequencer the group debuted two years ago, the new instrument, to be completed this fall, will be cheaper to buy and operate than competing next-generation sequencing platforms, and its throughput will be greater, the scientists claim.
While the developers currently use their own polony sequencing-by-ligation chemistry, the system will be open, allowing users to run any sequencing chemistry they choose.
The new instrument, which unlike its prototype will come pre-assembled, is “designed to be more reliable, more inexpensive, and faster” than its predecessor, according to Greg Porreca, a postdoctoral researcher in the Church lab. “We expect very little technical expertise will be required to keep it running well.”
“The price tag definitely gets my attention,” said Kevin Knudtson, director of the DNA facility at the University of Iowa. “This could allow some sites that may not be able to afford the commercial platforms to participate in this technology.”
About six months ago, Church’s lab hired Richard Terry, who has a background in mechanical and aerospace engineering, as a senior research engineer with the mission to design the new platform. “Basically, we went about redesigning the entire system from scratch,” Terry said.
“It’s designed for robustness, ease of use, speed, simplicity, and manufacturability,” he added, with the goal “to manufacture hundreds of these things in short order.”
An important decision, already made for the prototype, was to use the fastest and most sensitive camera available, Porreca said, so the detector would not limit the system’s performance. At about $30,000, the camera is “by far” the most expensive piece in the system, he added.
The researchers currently estimate the total cost at $95,000, a price that might still change by 10 percent or so in either direction. Free software to run the instrument is included, but data storage is not part of the package. No additional equipment other than standard molecular biology equipment is required for sample prep.
The new system includes a one-megapixel camera with a Nikon objective, a fast linear motor-driven stage, and an array of up to 36 disposable flow cells that can be run in parallel (for further technical details, see sidebar). While one half of the array is being imaged, the other half undergoes sequencing reactions, a design feature that both Applied Biosystems’ SOLiD and Helicos BioSciences’ HeliScope use to increase throughput. For some components, the researchers have contracted development to their vendors.
A run on the instrument will take “a few days,” according to Porreca. Each flow cell holds a chip with approximately 50 million beads, from which the researchers currently obtain between 5 million and 10 million placeable reads. Using all 36 flow cells and the Church lab’s four-color sequencing-by-ligation chemistry with its 13 base-pair reads, the system could thus produce up to 4.7 gigabases of sequence data per run.
Illumina’s Genetic Analyzer, by comparison, currently generates one gigabase of data per run, which the company said will increase to more than 3 gigabases by the end of this year. ABI has said its SOLiD platform will churn out up to 1.6 gigabases from a mate-pair library this summer, and 2-4 gigabases later this year.
With the current SBL chemistry, the median sequencing accuracy of the prototype is on the order of 99.7 percent, and the consensus error rate is less than one per million at 3X-4X coverage, according to Porreca.
But the new platform is not confined to the short-read SBL chemistry. “Our instrument, in theory, can implement any sequencing biochemistry you want to put on it, because it has temperature control and you can pipe as many different reagents in as you want,” Porreca said, adding that the researchers are considering several options to get longer reads.
For example, they could improve the ligation-based chemistry, or use a polymerase-based sequencing-by-synthesis chemistry to generate longer reads. Porreca said he is not aware of any sequencing chemistry — with the exception of pyrosequencing, which requires real-time monitoring — that is incompatible with the system. The instrument allows scientists to use any wavelength, and create new scripts for their own biochemistry. The researchers do not know if the sensitivity would be high enough for single-molecule sequencing but said that the optics could be easily changed.
One company that has been developing a sequencing-by-synthesis method is Intelligent Bio-Systems, which licensed the technology from Jingyue Ju’s lab at Columbia University last year and plans to ship early-access instruments by year’s end. George Church joined the company’s scientific advisory board in January.
Asked whether the company is collaborating with Church’s group, IBS CEO Steven Gordon told In Sequence last week that “We have talked somewhat about cross-pollinating, but at this point it’s separate platforms.”
For the Harvard researchers, what really counts is the sequencing cost. Throughput or bases per run are “absolutely a meaningless statistic,” Porreca said. “Really, what we are interested in, and what we have been interested in from the beginning, is cost per base,” he added. “That’s what anybody really cares about.”
Porreca hesitated to estimate the sequencing cost on the new system “until we have everything running.” However, he said that the cost — which will be largely a function of the reagents used — has already come down from 8 cents per kilobase in 2005 to around 2 cents per kilobase, or $20,000 per gigabase, running the polony SBS chemistry, which uses off-the-shelf reagents.
“And I would expect the new instrument to drop us down by another factor of somewhere between four and 10 from that,” he said, thus possibly decreasing the cost of sequencing to $2,000 per gigabase. Instrument amortization is only a small part of the cost, he said, because of the high throughput.
For comparison, Illumina cites $3,000 in reagent costs per run, which currently generates 1 gigabase of data, and ABI has quoted $3,000 in reagent costs per gigabase for its SOLiD sequencer, depending on the application.
The Harvard researchers are hoping to receive the first production model in their own lab in October and are planning to test it in-house “for a short period” to make sure the software and process are working properly. Within months, they plan to order more instruments for their own lab, “and at that point, outside researchers will be able to do the same,” Porreca said. Several labs have already expressed an interest in the instrument, he added.
The suppliers of the instrument’s parts will be prepared for what the researchers believe might be high demand. “These vendors, including the camera vendors, … can [provide] hundreds of units at a time,” Terry said.
The instrument will come from one supplier and require little or no assembly, he said, although researchers will also be able to access the original blueprints. Most likely, one of the “many” suppliers of the system’s components will handle orders but “nothing is written in stone” yet, he added.
As a PhD student in the Church lab, Porreca helped develop the prototype polony sequencer that the researchers published in Science in 2005. Since then, the lab has built several additional instruments that were “pretty much copies of the original,” he said. The prototype “was built with standard, off-the-shelf components that were not designed for the purpose,” he added.
One of the biggest problems with the instrument was that it required expertise to assemble it from separate pieces. Porreca knows of two laboratories outside of the Church lab who have adopted it: Jon Seidman’s group at Harvard and Jeremy Edwards’ lab at the University of New Mexico. Three weeks ago, Seidman’s group published an article in Science, using the prototype to sequence mRNA in heart muscle. That study included various improvements on the polony SBL technology (see In Sequence 6/12/2007).
“There is a perceived barrier of entry with the current prototype, and that, for the most part, disappears with the new instrument,” Porreca said.
In anticipation of Church’s budding personal genome project, during which he plans to sequence large numbers of individuals, the lab decided to build an improved version of the instrument that is cheaper and faster, and make it widely available to other researchers.
Still, some academic researchers are concerned about the level of support they would obtain from a non-commercial development group, or about the extra costs for support or additional equipment for data storage.
The new instrument is “designed to be more reliable, more inexpensive, and faster” than its predecessor.
According to Porreca, customers will get some support from the Harvard researchers, who are trying to build a “web-based community” to answer questions and help with troubleshooting.
The researchers also have faith that their instrument design will curb the need for support. “The instrument is really the most difficult part for the consumer to handle,” Porreca said. “So to the extent that it has been designed and implemented properly, a lot of the support, we think, goes away.”
As the need grows, third-party providers could also offer technical support for hardware and software, “and there are companies who have voiced interest in that,” Terry said.
Also, the researchers hope that users will help develop new sample prep methods and improve the sequencing biochemistry. “These are areas traditional DNA sequencing technology consumers are comfortable with,” said Porreca. “We hope that by placing a high-quality, efficient, open-source instrument on the market, we will enable all sorts of both large sequencing projects, at low cost, and development to improve the methods.”
Users will also have “unlimited flexibility in designing their own application-specific [software],” he said.
Porreca sees a number of applications for the instrument. “I would say expression profiling today, and I would say exon sequencing tomorrow,” meaning within a year. “Really, when you talk to most people, what they are interested in is human sequence, to whatever extent you can afford to produce it,” he said.
Academic researchers, polled by In Sequence, said they like the idea of a low-cost alternative to the high-throughput sequencers provided by 454, Illumina, and ABI but voiced some concerns.
“Anything coming out of George Church’s lab is worth looking at in detail,” George Grills, director of operations of core facilities in the life sciences and of advanced technology assessment at Cornell University, wrote in an e-mail message. He stressed, though, that researchers would need to see and compare data sets from this and other next-gen platforms.
Grills and others also wondered how easy the instrument and its software will be to set up and run, what type of customer support will be available for hardware and software, and how many different applications the platform would support.
“That being said, we expect to need additional next-generation sequencing platforms at Cornell, and we would definitely consider buying such an instrument,” Grills said.
“We will trade a lot of convenience, i.e., ready-made kits, for lower cost,” said Bill Farmerie in an e-mail. He heads a core facility at the University of Florida, Gainesville, which owns a 454 sequencer.
Regarding the instrument price, he said that although it is a high upfront investment, “eventually, the cost of operation — labor, reagents, and supplies — dwarfs the initial cost.” Reagents contribute most to operational costs, and “perhaps an open platform not tied to a particular vendor chemistry creates healthy competition, drives down cost, and/or stimulates development,” he said.
Support for an open platform would depend on the community of users, “so success depends on sharing information and experience,” Farmerie cautioned, while noting that this model has worked well in the past.
A single vendor who also provides support for a system worldwide “may also identify problems more quickly than a highly decentralized instrument and supply chain,” he said.
“I would definitely consider acquiring this instrument — although I’d need to know more about it,” he said.
Several researchers who are interested in de novo sequencing as well as other applications said they do not like the short read length of the SBL chemistry.
“We could use the approach for resequencing and SNP analyses, but read length would be too short to contribute significantly toward de novo sequencing,” Loren Rieseberg, a researcher in the botany department of the University of British Columbia in Vancouver, wrote in an e-mail message. Rieseberg won 100 megabases worth of free sequencing from 454 earlier this year. Nevertheless, he said “I think we could find many uses for the instrument.”
“In a core setting, I’m interested in something that has versatility,” including the ability to be used for de novo sequencing, said Knudtson from the University of Iowa. He also wondered how much data storage equipment would add to the price tag.
According to Terry, data storage and analysis equipment could range from a couple of thousand dollars to hundreds of thousands, depending on how much data researchers want to store for how long and at what rate, and how much they want to process it.
“If it works, and the sequence length and quality is good, and the price of chemicals to run it is not too high, I would guess many labs or universities would be interested,” said Kris Lambert, an associate professor in the department of crop sciences at the University of Illinois at Urbana-Champaign, who attended a company presentation on ABI’s SOLiD sequencer earlier this year.