Under the Hood with Edico’s DRAGEN Card at HudsonAlpha
By Allison Proffitt
September 18, 2015 | Shawn Levy isn’t interested in arguing about the numbers. He’s content to say that HudsonAlpha is one of the largest sequencing centers, and with a staff of just 10—including himself and the office administrator—he leads one of the most efficient teams.
HudsonAlpha’s sequencing center is home to nearly the whole Illumina product line: 8 HiSeq Xs, seven HiSeq 2500s, five HiSeq 2000s, a couple of MiSeqs, a NextSeq, “and a partridge and a pear tree,” Levy jokes. He’s been running the Genomic Services Lab at HudsonAlpha for six years.
Right now the center is sequencing a little more than 15,000 genomes each year—at current volumes, Levy expects his annual number this year to be between 15,200 and 15,500. Customers include 700 labs around the world, “three major pharma, and lots of medium, though some of the medium are growing really fast,” Levy said. “If you count a penguin microbiome project, we’ve hit all continents.” The Genomic Services Lab’s second experiment will go into space next year.
Levy’s main priorities are “ridiculously high expectations for data quality”, efficiency of process, and a financially-responsible model for a genome center. “One of these is forced,” Levy said of the financial responsibility. “The other two are the way it should be.”
“Like many centers,” Levy told Bio-IT World last month, “we’ve always lagged behind in our ability to analyze and store data compared to how fast we can produce it.” At HudsonAlpha, he said, he has “exceptional freedom” to fix problems. While there are dedicated IT teams committed to fixing data analysis pipelines, Levy still needed to “put a Band-Aid” on the system so it could continue to deliver data, “at the expectations we’ve set both for ourselves and our collaborators. So we built an alternate infrastructure that includes the Edico DRAGEN card.”
Edico’s DRAGEN card made waves last year with its claims of speed—it was a finalist in the 2014 Bio-IT World Conference & Expo Best of Show competition and announced its first customer one year ago.
But Levy is not one to buy into marketing claims. “As it comes from Edico, it’s just an appliance,” Levy notes. “It doesn’t have any front end in front of it that actually feeds it. It’s like if you thought you were buying a car, but the car manufacturer delivered you an engine.”
But don’t let Levy’s candor suggest he’s not pleased. “It’s an amazing appliance,” he said. “We’ve worked closely with them,” he added. “It’s actually a pretty fantastic appliance
The DRAGEN genome pipeline includes highly optimized algorithms for BCL conversion (the company claims to do the BCL to FASTQ conversion in 12 minutes), compression, mapping, alignment, sorting, duplicate marking, haplotype variant calling and joint genotyping, and the reconfigurable processor can be loaded with additional pipelines, such as RNAseq, methylome, microbiome and cancer. DRAGEN is available in a pre-configured server that the company says is easily integrated into next-generation sequencing bioinformatics workflows.
HudsonAlpha developed a front-end management system to enable robust and automated transfer of data from a collection of sequencers to the processor, and automated reporting of results at the conclusion of the analysis. The supporting workflow that Levy’s lab developed will now be freely available to academic labs.
Edico claims that DRAGEN is able to “analyze a whole human genome (30x coverage) from FASTQ to variant call format (VCF) in under 28 minutes, compared to over 14 hours using standard software.” Levy has found that a single DRAGEN system will keep up with an Illumina X10 and can process a genome every 40 minutes. And again: “It’s fantastic!”
DRAGEN is currently used in Levy’s Genomic Services Laboratory, as well as in its newly established Clinical Services Laboratory, which performs whole genome sequencing for patients with medical conditions. The processor also is used in HudsonAlpha’s Clinical Sequencing Exploratory Research (CSER) consortium project, initiated by the National Human Genome Research Institute (NHGRI) and the National Cancer Institute (NCI) and uses exome sequencing to diagnosis children with developmental delays.
Levy said HudsonAlpha found itself on a “collision of circumstances” that required far more compute than previously estimated. “It was an underestimate on our part as an institution of how much compute we’d need at a given time,” he said. “And since we realized that all at once, we thought, ‘What’s the solution?’ And we brought on the Edico card.”
Levy credits the system with preserving HudsonAlpha’s ability to meet its commitments. “The implementation of that DRAGEN card and the speed with which we’re able to do it completely saved our ability to deliver genomes on the timescale we were promising,” Levy said. “Without them we would have been in terrible shape.”
And others at HudsonAlpha are eager to try it as well. Liz Worthey, who with Howard Jacob and others moved to Alabama this summer from the Medical College of Wisconsin, told Bio-IT World that she is looking forward to seeing, “if the DRAGEN card can be clinicalized.”