To Speed Sequencing, Edico Genome Proposes NGS Processor
April 21, 2014
By Bio-IT World Staff
April 21, 2014 | Of sequencing bottlenecks, the most lamented is the analysis pipeline. Edico Genome is developing the DRAGEN Bio-IT Processor, the world’s first Next Generation Sequencing (NGS) bioinformatics ASIC to speed and smooth that process. In alpha testing now, the processor stands to revolutionize genome analysis, Gavin Stone, Edico’s Director of Marketing told Bio-IT World.
Of course, that’s language you’d expect from a director of marketing. But early feedback looks promising. The company’s advisory board boasts Eric Topol, M.D. (Scripps), Charles Cantor, Ph.D. (Sequenom) and Nils Homer, Ph.D. (Broad Institute). Alpha customers are already using the product.
Bio-IT World’s Best of Show judges are eager for a closer look. The product is a finalist in the Best of Show competition and the company will be exhibiting at the Bio-IT World Conference & Expo*.
Edico, “essentially takes the data that comes out of a sequencing instrument, and we process that on a hardware chip—which is completely different from how everyone else is doing it in the industry, when they just throw a cluster of compute servers at it,” Stone says.
As a case study, he uses Illumina’s new HiSeq X Ten system. The system, “can sequence 150 whole genomes every three days. At the end of each round, you’ve got 150 whole genomes worth of data that you have to analyze.” If you want to keep up, Stone argues, you should analyze those 150 genomes within three days, before the next wave of data comes off the system.
“You’ve got three days to get through 150 whole genomes, which means you’ve got 28 minutes to get through the analysis of a single genome. But to do all that, you need around 50 high end compute servers… at a significant cost to buy the equipment, run it, power it, have IT staff to manage it. We essentially replace all of that with one of our cards.”
Edico estimates the savings at $6 million over four years.
The DRAGEN Bio-IT Processor is integrated on a PCIe form factor card and is provided with accompanying software as a Platform-as-a-Service (PaaS) platform that can be integrated into OEM sequencing machines and NGS bioinformatics servers. The processor is loaded with proprietary algorithms for mapping, alignment, sorting, and variant calling.
Edico’s analysis pipeline consists of, “our own variations of algorithms like BWA and GATK, which typically aren’t suited to implementing on hardware,” Stone said. “We’ve done a hardware-specific optimization and implementation of those algorithms.”
The result, Stone said, is “massive speed and throughput benefits.” A typical whole genome mapping, aligning, and variant calling with BWA and GATK takes 20-24 hours to do with a top of the line server. Stone says Edico can do it in less than 18 minutes.
But Edico isn’t sacrificing accuracy for speed. He claims the company is seeing improved accuracy with standard algorithms. “Because we have just a massive, sheer force of putting everything on hardware, we don’t have to compromise on anything,” Stone said.
For example? Stone says Edico has the processing power to use a “full Smith-Waterman array” for alignment, whereas software often “makes guesses and takes shortcuts” to create alignments.
We asked Stone what Bio-IT World attendees can expect to see from Edico at the end of the month. He said the company plans to launch their availability program, demonstrate the processing power of the DRAGEN Bio-IT Processor compared to BWA, and present the HiSeq X Ten case study.
The company also plans to invite Bio-IT World Expo attendees to evaluate the speed and accuracy improvements afforded by DRAGEN Cloud, an online version of the DRAGEN Accelerator Card. Users will be able to upload a gzipped FASTQ file, run it through DRAGEN Mapping/Aligning/Sorting Pipeline, and then download the BAM file for further analysis and comparison with results from existing solutions like BWA.
*Editor’s Note: The Bio-IT World Conference & Expo will be held in Boston on April 29-May 1. For more information, see www.bio-itworldexpo.com. Edico Genome will be exhibiting at booth #221.