Because marketing is not data. Businessweek is trumpeting Oxford Nanopore’s “unveiling” of new sequencers at the American Society of Human Genetics meeting (boldface mine):
Oxford Nanopore Technologies Ltd., the closely held company developing a portable gene-sequencing device that will sell for less than $1,000, plans to unveil two instruments that use its novel technology today.
Oxford Nanopore will show the portable product and a desktop machine to doctors and scientists attending a conference of the American Society of Human Genetics in San Francisco this week, Chief Executive Officer Gordon Sanghera said. He declined to specify when the company would begin taking orders, saying that it plans “commencing commercialization” this year….
Sanghera declined to disclose the speed or capacity of the MinION and GridION machines and wouldn’t confirm whether they are being field-tested by prospective customers. The company said in February that the devices had an accuracy rate of about 96 percent, comparable to machines on the market, and said it wouldn’t sell the products until they were accurate more than 99 percent of the time.
“I can’t comment if we are there yet,” Sanghera said.
Before I tear this a new hole, let me be clear: I want Oxford Nanopore to work. If it works as advertized, it will increase speed, lower computing costs, and decrease personnel costs (though high-quality de novo bacterial genomes, at least in my neck of the woods are already really cheap). But the Businessweek story is setting off all sorts of bullshit detectors. First, a 96% accuracy is “comparable to machines on the market”, if by comparable you mean about four times as worse. It still sounds like they haven’t solved the insertion/deletion (‘indel’) problem, which means that annotating genes (figuring out what genes the DNA sequence actually contains) will be a nightmare–if you lower sequencing costs, but increase human analyst costs, that’s not an improvement: What profit a Mad Biologist who gains cheap sequence only to have to pay tons of money on the back end? (I think that’s in the Bible somewhere).
If they had solved the indel problem and produced a kickass sequence of, let’s say, an E. coli K-12 genome (a standard test organism for sequencing technology), wouldn’t they have released the data? Hell, wouldn’t they be shouting it from the rooftops? So where are the data?
The vagueness sounds like a hybrid of ass-covering and marketing hype, but I hope I’m wrong.
mikethemadbiologist
The lack of data is a bit worrying. However, the indel problem (or any random error) is easily circumvented with a decent reference genome coupled with sufficient over-sequencing. I therefore think it will be initially be useful for resequencing projects rather than de novo.
There will also most likely be whole new suites of analysis tools in the pipeline that will allow us to make use of this data. Most of the current aligners would struggle with the 40kb read length now quoted. I would be surprised if there wasn’t some way of identifying these errors (lower Q-scores?)
I honestly believe Oxford nanopore is beginning with the left foot here. All the announced “disruptive” technologies in the last couple of years haven’t been doing so well (with the clear exception on ION torrent) and they seem to be drawing the same path. I hope they succeed (sooner or later) but these last lack of results have been disappointing for many. Maybe when they can show some real results, it might be too late…
We absolutely need to see data, but long imperfect reads are fine for de novo assembly, if you combine them with short lower error reads (Illumina or Ion Torrent). This combined approach is straightforward & what’s being done with PacBio data. I know Oxford would rather users relied upon their data entirely, but why not start with something that’s an immediate help to many of us & could let us stop making mate-pair libraries evermore?
We’ve been begging for a handful of mini-ions to test with various sample types, and various prep ideas… no dice. I even asked for some specs, like expected input amounts and concentrations, so we could “get ready” …. no response. I think Oxford is wasting a lot of time developing “friendly software”, which is exactly what PacBio did. I hope I don’t have to fight with their software to get “basecalls and quality scores” … sounds like it though.
Pingback: My take on the sequencing buzz at #ASHG2012 « In between lines of code
I’m hoping they’ll ultimately pull a Steve Jobs on us when they do announce: “We said you’d get 100,000 bp long reads with 96% accuracy and people have been doubting that… in fact you’ll get 1,000,000bp long reads with 99.9% accuracy”
Pingback: Links 11/20/12 | Mike the Mad Biologist