3rd-generation Sequencing: A RISC-y Business
18 May 2010Nature has a useful summary of single-molecule or '3rd-generation' sequencing technologies based on information released at AGBT earlier in the year. Most of this was already known thanks to the reporting from bloggers like Dan Koboldt, Luke Jostins and Daniel Macarthur at the time, but the Nature article is sure to bring 3rd-generation to wider attention.
Locally, we're considering the purchase of an Illumina HiSeq 2000. This represents the very cutting-edge of the 2nd generation technologies, with throughput around 200-300 gigabases per run. But a niggling thought won't go away: what if 3rd-generation instruments come and blow even this technology out of the water, and soon.
So very similar to the common nerd dilemma: buy a laptop now or wait for the next model?
The well-worn analogy is with Moore's law, we only need to wait 18 months or less before a model twice as capable is released. But does this hold for the transition from 2nd to 3rd-generation sequencing? And indeed, is such a transition inevitable? The newcomers to the scene will tell you it is.
Continuing the analogy with silicon chips, it occurred to me that a useful lesson from history might be the hype associated with RISC chips in the early 90s. You'll remember that the hype around RISC was intense back then. It was going to kill off the dominant Intel x86 (8086, 80386, Pentium etc.) processor family for good. Apple bet the farm on IBM's PowerPC chipset for its new Macs, probably the wrong decision and they eventually moved back to Intel. Check out this classic advert.
Despite the seeming obvious killer advantages over CISC chips - lower power, faster, simpler to manufacture, RISC chips resoundingly failed in the desktop PC market, never challenging Intel's dominance.
I propose that Illumina are Intel, and the Genetic Analyzer family (GA1, GA2, GA2x, HiSeq 2000) are x86. Life Tech is AMD, producing similar technology with much reduced market share. And tThe 3rd-generation technologies could end up repeating the RISC story.
Let me explain myself. And I would like to point out, I am as excited as the next man about new sequencing technologies. I would like the innovation in this space to continue forever. I am talking about market forces as much as I am about technology.
Wikipedia reckon there are 3 reasons that RISC failed to unseat Intel's CISC-based technology.
Firstly, when RISC came out, there was a huge base of proprietary applications available that only ran on Intel x86 architecture. This gave Intel a huge advantage. Developers would have put a lot of effort in to develop for RISC architectures to get the same stuff running.
We see the same phenomenon in the second-generation space. Both the academic community and the commercial companies have invested in the 2nd-generation space. Large numbers of de novo assemblers and aligners have been developed specifically for Illumina data. We've already seen that much fewer software is available for SOLiD's colourspace data. Will cutting-edge software be developed for newcomer 3rd-generation sequencers or will this be left to the instrument manufacturer? Biotech hardware manufacturers like Agilent and Covaris are now making hardware specifically to help speed up and automate the Illumina workflow. Enzyme producers such as New England Biolabs are developing novel enzymes for the Illumina workflow. Knowledge is important too: people are now comfortable with the relative foibles of 2nd-generation machines: short-read and low-quality tails in Illumina data, homopolymers in 454.
Then there's the Intel factor. A major reason the x86 platform prevailed was their market dominance. An aggressive and single-minded strategy focused on rolling out regular and predictable improvements to the x86 chipset, despite the seemingly unsurpassable obstacles in their way. This was possible due to the huge revenues they were pulling in from being the dominant chip manufacturer. By aggressively increasing processor clock speed and increasing the numbers of transistors on the chip, they managed to outgun the theoretical advantages posed by RISC chips. When they hit the power ceiling they resorted to adding more cores. This is very similar to how Illumina have juiced the Genetic Analyzer since 2007. Better imaging and processing has meant higher density clusters. Now with the Hiseq they process multiple flow-cells in each run.
Finally, Intel managed to outmanouvre the threat from RISC by copying some of the best ideas from RISC and integrating them into the x86 family.
We already know that Oxford Nanopore have signed a marketing deal with Illumina. It is always assumed that they will bring out their own instrument, but will we instead see this technology come as a bolt-on upgrade for a future version of the HiSeq?
And we've already seen the first single-molecule technology falter: Helicos have had a torrid time, failing to get any kind of market penetrance, and today announced they are laying off 40 staff.
If 3rd-generation tech is going the way of RISC, what predictions can we make?
Some, if not the majority, of 3rd-generation technologies will never make it to the market-place.
Certain 3rd-generation technologies will be successful but not in direct competition with Illumina. For example the Ion Torrent seems to be targeted cheap, fast, low-throughput sequencing. They may succeed in applications like clinical diagnostics in microbiology and the near-patient testing market where fast run-time is key. Think Star Trek style devices which you wave over the patient. The iPhone of sequencing. In fact this platform comes with an iPhone so they are obviously thinking along the same lines ;)
I think that for sequencing applications where throughput and price is the primary concern (say, human resequencing for SNP detection), Illumina will continue to dominate the marketplace for the next 5 years with the Genetic Analyzer, despite several 3rd-generation technologies being mature by then.
What do you think?
Conflict of interest declaration: Would love an Ion Torrent, Pac Bio or Oxford Nanopore system but so far have been unlucky in my attempts to get one! ;)