This story was originally published May 23.
Researchers at MIT have developed a method to give aging Illumina Genome Analyzer II sequencers a second wind by running the lanes of a flow cell separately.
While many laboratories have already abandoned their GAIIs in favor of Illumina's newer models – the HiSeq and MiSeq – the new protocol, published in BioTechniques this month, might allow those still hanging on to their machines to get more mileage out of them. In addition, similar modifications might work for the HiSeq, increasing its flexibility.
According to Stuart Levine, director of the BioMicro Center, a core facility at MIT that serves several departments and centers, and the senior author of the paper, the method extended the life of the three GAII sequencers in his lab. "If I didn't have it, I would not be using my GAIIs. I'd be using them almost never," he said, "and now I'm using them a couple of times a month at least."
The BioMicro Center is also equipped with a HiSeq 2000 and a MiSeq, which have taken over much of the usage for the GAs. To enable them to run smaller-scale experiments on the instrument, Levine and his colleagues developed a scheme to run partially loaded flow cells on the GA without wasting reagents or remaining lanes, which they started using in their core about a year ago.
Both the GAII and the HiSeq have flow cells with eight independent lanes that usually run in parallel, so core facilities often have to wait for enough samples before they can start a run. "We were not getting enough samples to fill flow cells" for the GAII, Levine said. "If people wanted to run the GAII for any number of reasons, rather than the HiSeq, we needed a way for them to be able to do that, particularly without waiting months for other people to drop off samples."
In order to run the lanes of a flow cell separately, and to use the remaining lanes at a later stage, the researchers made several "fairly straightforward" modifications, both to the hardware and to the fluidics scripts that run the instrument.
To prevent reagents from being pumped through certain lanes, they disconnected unneeded syringe pumps by unscrewing the plunger heads from the crossbar that moves the syringes, a procedure that only required taking off the housing and using a flathead screwdriver.
To grow DNA clusters in only some of the lanes, they cut off unneeded cluster reagent tubes using a razor blade, replacing them with tubes containing high salt buffer.
Finally, in order to correct for dead space in the lines, they adjusted the reagent volumes that get pumped through the system, writing a script that changes the XML scripts.
"I think it's pretty easy," Levine said, especially using their script, which automates the reprogramming. "There is nothing here that's rocket science."
"Most labs have a flathead screwdriver, and the type of people who run cores typically are the type who like fiddling with things," he said.
The researchers validated their method by running a control library in two lanes of a flow cell and found that the error rate stayed below 1 percent up to a read length of 80 bases and climbed rapidly after 100 bases, more than during a standard, full flow-cell run.
They also ran the same flow cell five times, using different lanes in each run, and found that the per-base intensities did not decrease significantly in later runs.
In addition, they ran a paired-end protocol, generating 2x40-base reads, on four lanes and found that the error rates did not differ much from those of a full run.
Over the past year, Levine's lab has used the lane-by-lane method extensively. "Almost everything we do on the GA now is partial flow cells," he said.
One application is paired-end runs that require a different reverse read primer for each lane, something that is impossible to do in a standard GAII run, where each lane receives the same reverse read primer.
They have also used the method for high-throughput sequencing fluorescent ligand interaction profiling, or HiTS-FLIP, which measures protein-DNA binding affinity, because they can deliver a different protein to each lane.
The BioMicro Center's website recommends the GAII for runs with unusual read lengths, protocol prototyping, and non-standard assays, such as HiTS-FLIP.
"Whenever we have some weird things that people are doing, we're pretty much always pushing them onto the GAII because we have a lot more flexibility of what we can load … we have a lot more control," Levine said. "If it's really experimental, it's often a lot easier to run on the GAII."
According to the website, Illumina does not support the partial flow cell method, so the center does not offer customers to repeat failed runs for free, even if they were caused by faulty Illumina reagents. However, Levine's team acknowledged two Illumina employees in its paper who he said helped with some technical challenges in developing the method.
Jeremy Preston, director of product marketing at Illumina, told In Sequence that Illumina "only supports validated workflows and protocols" and has developed dual lane flow cells for the HiSeq 2500 that "address the need to run lower numbers of sample per run in a rapid mode configuration." He added that Illumina is unlikely to pursue further developments on the GA.
Levine believes that the method can in principle be extended to the HiSeq, which he said uses the same basic infrastructure as the GAII − a crossbar to move the syringes and XML scripts – but he has not had a need to run partial flow cells on the HiSeq because the runs are "pretty close to full."
"If somebody has the resources, the ideas are there. It's very doable I think, but I can't be the one to do it," he said.
It is unclear how many GAIIs are still in use, or how long Illumina will continue to support the platform. "I assume a fair number of them [are still out there]," Levine said. "They're very cheap to buy right now."
Preston said the company does not comment on the size of its installed based but noted that it has "a sizeable fleet of GAII machines regularly running across the world."
Vladimir Benes, director of the genomics core facility at the European Molecular Biology Laboratory in Heidelberg, Germany, said his lab gave up its last GAII nearly two years ago. He thinks the new protocol is clever but probably has limited utility. "It is not trivial and requires a lot of preparations before you can run [the GAII] in the lane-by-lane mode. People who haven't, or couldn't, use their GAIIs fully up to now wouldn't go for it," he said.
Levine thinks otherwise. "We're really hoping other people will use [our method] and bring the GAIIs back to life a little bit, and possibly even improve on it," he said, for example by increasing the base accuracy and extending the usable read length.