CHICAGO (GenomeWeb) – Optalysys, a British company developing high-performance computing systems based on cost-saving optical technology, is just about ready for a commercial release of its technology for DNA sequencing alignment.
Optalysys will announce details of its first commercial product in early 2018 after testing its final prototype, according to CEO Nick New. That work should wrap up by December.
"The last tests and stages of the project involve running the system at speed. We are just awaiting delivery of the drive electronics, which has been a major undertaking for us," New said last week.
There has been a delay waiting for manufacture and shipment from the US of some custom electronics for the final prototype of the technology, according to Timothy Stitt, formerly head of scientific computing at the Earlham Institute, which has set the scientific direction of the project. "We can't do our final performance testing and energy-efficiency testing until those are in place," noted Stitt, who became head of HPC strategy and architecture for pharmaceutical giant Roche this month.
The Norwich, England-based Earlham Institute, known until June 2016 as the Genome Analysis Centre, is a genomic sequencing, computational biology, and research center.
In 2015, the institute tapped Yorkshire, England-based Optalysys to develop a commercial platform for searching DNA sequences with optical processing technology. Innovate UK, the British government's official innovation agency, gave TGAC and Optalysys a two-year, £500,000 grant (now worth about $679,000) for the project.
The system couples Optalysys' optical technology with Blast and BWA software to enable researchers to run large-scale DNA sequence searches without the need for expensive, energy-hogging high-performance computing systems. The project is aimed at shrinking a 28-hour process to 1.6 hours, while slashing power consumption by 90 to 95 percent, according to New.
Because it's optical, the technology does not generate heat and require cooling.
"With this technology, you just attach it to a regular desktop PC. You don't need a large data center or large power requirements. You just power it from a standard wall socket," Stitt said.
"This device, we reckon, is just going to require 50 to 60 watts of power," Stitt added. Earlham's existing HPC resources can consume as much as 130 kilowatts of power, including cooling systems to remove heat.
Earlham and Optalysis haven't hit those goals just yet, but they have been successful in replicating the accuracy of HPC for sequencing alignment.
Last month, Daniel Mapleson, Earlham's lead bioinformatician for the project, presented preliminary results at the Genome 10K and Genome Science 2017 Conference that Earlham hosted in Norwich. In examining a 20 percent mutation rate of the E. coli K12 genome, the Optalysys optical correlator was just as accurate as a Blast installation on HPC infrastructure, according to a poster that Mapleson presented.
"We expect further improvements to accuracy by refining the optics and optimizing the software," the poster said.
"Most of our focus within the last 12 months has been on the accuracy side of things," Stitt noted. Accuracy means finding alignments that BWA and Blast are finding, as well as other alignments that those platforms might miss, according to Stitt.
"There's no point in us building a really energy-efficient device that's really quick but gives us worse results than if we just ran it on an HPC system," Stitt said.
Once the custom components arrive, the team can concentrate on speed and power consumption. "The final stage of [testing] is to then raise the speed of the system to accomplish the goals of the project. For that, we need the drive electronics to drive our optical processor to reach its potential," said New.
"Within the last few weeks of the project, we'll work on the performance and the energy efficiency once we have the final prototype in place," Stitt said. He expressed confidence that they will meet the energy goal. "We're also hoping to be much quicker ... than running Blast or BWA on an HPC platform," he added.
A "future development" is to remove the requirement of a large HPC setup and perform genome alignment on a desktop PC, New said.
"The scientists, the bioinformaticians, they get their results in a much faster time," New said. The work can be done within the lab at a speed comparable to an HPC node. "You're taking a software[-based] electronic process and running it through this optical process. You can produce those results in a much faster time scale."
That would, in New's estimation, help Optalysys shake up the marketplace for sequencing alignment. He said that even those with HPC installations would want to use it because of the speed and power efficiency.
"With this new technology, we can do this really, really fast and also do it very, very energy efficiently ... which is where some of the big HPC platforms are starting to struggle." Stitt said. This makes sequencing alignment accessible to smaller institutions.
Since the project started in 2015, Optalysys has learned different ways of encoding data within the device, according to Stitt. "They've come up with some really clever ways of mapping this sort of traditional format of DNA sequencing data into the device as optical patterns. They've come up with some really clever mathematics to push through as much data as they possibly can through the system," he said.
While the exercise originally was only supposed to run Blast software on the optical processor, the researchers also looked at BWA. "We can actually do both types of alignments on the same device," Stitt said.
That is not possible with standard HPC setups. "That requires learning two different software tools, learning the best practices for each to get the most efficiency and productivity out of them. With this device, we can now do both alignments within the same piece of technology," Stitt said.
"We really want this technology to be picked up by the bioinformaticians in the community. The more capabilities it has, the easier it makes life for the bioinformaticians," he said.
Future releases likely will include other types of alignments, including RNA-Seq, according to Stitt. "It's just one device required to do all of this, so hopefully that will make it attractive to bioinformaticians," Stitt said.