By Joe Stanganelli
September 30, 2016 | In the world of next-generation sequencing (NGS) research and diagnostics, 2016 is shaping up to be a big year for the Broad Institute's Genome Analysis Toolkit (GATK). Several announcements and releases this year have helped make the platform ever more ubiquitous as even Broad competitors have partnered up with the nonprofit for GATK compatibility—including Edico Genome's announcement this past May that the company will make an "accelerated" version of GATK available on its DRAGEN Bio-IT Processor.
Still, not everyone is inclined to agree that GATK is the NGS heir apparent. Indeed, the world-record holder for accomplishing the fastest genetic diagnosis has very different preferences.
For him, 2016 is the year of the DRAGEN.
"The Best of Both Worlds"
This past April, Stephen Kingsmore, President and CEO of Rady Children's Institute for Genomic Medicine, was officially acknowledged as having set the Guinness World Record in speed of genetic diagnosis—having used genomic sequencing to "successfully diagnos[e] critically ill newborns in just 26 hours". And the open-source, nonprofit-originating GATK appears to have had nothing to do with it.
Instead, according to Kingsmore, the record was made possible by a combination of Illumina technology and Edico Genome's DRAGEN—the latter of which Kingsmore reports being introduced to at last year's Advances in Genome Biology and Technology (AGBT) conference.
"At that point we were just finishing up some work to set the new world record for speed [of] turning around a clinical genome sequence, and we thought that [DRAGEN] would be a good thing to evaluate," Kingsmore told Diagnostics World News in a series of interviews. "We had been looking specifically at faster ways of computing a genome—and we had [previously] been looking at Illumina to do that; they had built a new algorithm which was very fast but was not very sensitive."
Kingsmore and his team ended up combining sequencing solutions from both Edico Genome and Illumina—using a methodology called STATseq, based upon Illumina's HiSeq 2500 system.
And for Kingsmore, two techs were better than one.
"When I heard about the Edico technology, I was a little bit skeptical, but thought, 'Well, let's try this out,'" related Kingsmore. "So we were able to send them some data sets and evaluate them together, and, much to our surprise, it is much faster than standard technologies and also retains full sensitivity—so it's kind of the best of both worlds."
Kingsmore went on to praise DRAGEN for being "particular[ly] sensitive for insertions and deletions" as well as for its overall "robust[ness]."
"Speed was what got us interested in the first place, but speed is something that by and large you use for world records and maybe not quite so much day-in day-out," said Kingsmore. "Initially it was about speed, but now we've really found that it's got multiple benefits for us.
The Impact of Velocity
Since deploying DRAGEN with STATseq – and subsequently setting the 26-Hour Genome record – Kingsmore and his team have fully deployed the technology for everyday differential-diagnostics use in neonatology and intensive-care for infants at Children's Mercy Hospital – San Diego. Kingsmore explained that the initial sequencing is, for the time being, done offsite (until Children's Mercy can get its own lab up and running) via collaboration with HudsonAlpha Institute subsidiary Envision Genomics. Envision in turn submits the raw sequence files back to Kingsmore's team via the DNAnexus cloud platform. The turnaround time from the moment the samples go out to Envision is about five days. From there, Kingsmore's team runs the sequence files through one of their two DRAGEN systems for alignment and variant calling, taking 40 to 50 minutes on genomes that are typically 40x to 50x. Finally, they use Omicia Opal and Qiagen's Ingenuity Variant Analysis for online clinical analysis and interpretation. Kingsmore reports that his team has done this for at least 14 families spanning a roughly 50-day period from about the second week of this past August through the last week of September—and the demand is ever present.
"Each week now, we're enrolling two or three families from our neonatal intensive care unit or pediatric intensive unit, and they undergo rapid genome sequencing," said Kingsmore. "It's our vision over the next year to scale this up to meet the needs of every child in our neonatal intensive-care unit [NICU], pediatric intensive-care unit [PICU], and cardiovascular intensive-care unit. We have regional centers for NICU and PICU [throughout San Diego County], and we have almost nine million people whose ultra-sick babies are getting transferred to our units—so you can imagine this is going to have a profound impact on care in San Diego County in southern California."
Money, Genes & Speed
That impact stands to be grander yet, if Edico Genome has any say in the matter. Kingsmore, while perhaps one of the company's more notable brand advocates (what with his DRAGEN-driven world record and all), is but one buck on the wall for Edico Genome.
"[We] have plenty of examples … where customers have switched to using DRAGEN to achieve the diagnostic goals of turnaround time or ultra-high throughput," said Gavin Stone, Vice President of Marketing at Edico Genome, in an email interview with Diagnostics World News. "The work that we have done with Children's Mercy Hospital and Rady Children's Hospital on the 26-Hour Genome is a great example of achieving something that was not previously possible in genetic diagnostics, [and] our work with HudsonAlpha on their high-throughput clinical sequencing lab really highlights DRAGEN's scalability."
Conversely, Stone expressed less enthusiasm for GATK.
"Unfortunately, I don't have any examples like this for GATK. Only a small percentage of our customers choose to license GATK from us given the performance that DRAGEN provides," said Stone. "We do however… offer standard GATK available on our DRAGEN platform that runs almost as fast as our DRAGEN genome pipeline for those customers that are locked into using GATK."
Indeed, while the Broad Institute makes GATK available for free to academic and nonprofit institutions, for-profit entities must pay a hefty licensing fee and, accordingly, potentially face the vendor lock-in issues Stone cites. Members of the bioinformatics and medical research community have roundly criticized the nonprofit for spoiling the open-source nature of GATK by using it to get into the commercial-licensing business with GATK—rather than keeping it entirely free, charging only for support.
"[C]ommercial-licensing revenue funds a portion of our GATK user support team, which right now is five people. The revenue raised doesn’t cover all of this, but it helps defray the costs," rebutted Geraldine Van der Auwera, one of the Broad Institute's Data Science & Data Engineering group leaders, via email on this point. "This is where it is important to note our status as a non-profit: revenue from commercial-licensing fees goes back into improving the service itself."
Stone, however, suggests that those improvements may not be worth the money.
"Our DRAGEN genome pipeline is an alternative to GATK that our customers are [choosing] to use over GATK given its massive speed improvement (~80x) and similar accuracy."
To speak further of money, Kingsmore, as a customer, has his own take.
"Traditionally, genome sensors have to have a supercomputer with maybe a thousand nodes in order to do alignment and variant calling, [but] DRAGEN is very cost effective from a hardware standpoint where a relatively inexpensive server with the DRAGEN card replaces a supercomputer—and that has profound implications for users," observed Kingsmore. "In addition, now, DRAGEN is going in the cloud, and so people will further be able to reduce their need for on-site computing, so those are very attractive things."
Kingsmore cited additional storage savings as well by virtue of, again, DRAGEN's sheer speed.
"One offshoot of the fact that it can process a genome really quickly is that we don't need to store BAM files … and the BAM is really big," said Kingsmore. "Typically with genome sensors you're spending millions of dollars a year just on storage, and DRAGEN means that you can cut that at least in half just by not needing to store BAM files. If you need a BAM file in the future, you just reprocess the genome—so there's no real need to store those."
"The Promise of Precision Medicine"
And then there is the matter of the positive ROI of speedy sequencing: more volume—those two to three families who get a genetic diagnosis every week.
"We've had a number of diagnoses and a number of babies, it's changed how we manage them in the intensive care units, [and] in a couple of the patients it changed the drug treatment that they would be receiving," noted Kingsmore. "This is the promise of precision medicine—that when we are able to make a molecular diagnosis of a genetic disease, that means we know how the disease is inherited. We know whether it will recur in subsequent kids. We have a good handle on what the prognosis is [and] what the complications might be, so we can predict that before it happens, and then we know the specific treatments that the child should get for their specific condition—and so in several of the cases, there have actually been changes to the drug that was given or the way that the baby was being treated … [including] decisions [related to] palliative care[.]"
And Kingsmore sees even greater potential in the form of scalability.
"It takes us about a week currently to get an answer back and then it takes us about another week to get a confirmed result, so it's roughly a two-week turnaround—unless the result is life-saving, in which case it's about a week; once we switch [our own lab] on here – the sequencing part – we'll be able to shorten that down probably to about three days," calculated Kingsmore. "And we're definitely not finished yet. The evidence so far is becoming more and more convincing that this is one of the first applications of genomic medicine that is ready for prime time and so it's very important for us to figure out how to scale this, [to determine] what … those protocols look like, and to then start to get that information out to NICUs around the country."