I read the coverage of a CNN piece on GenomeWeb with interest; the article talks about how much costs are dropping, but alternative views were only recently aired on GenomeWeb by Neil Hall and Mick Watson, so who's right?
Back in May a commentary article by Neil Hall and a blog post by Mick Watson, both discussed the very recent stop in the precipitous fall of sequencing costs. We've become so used to the continued fall that I suggested, very much tongue-in-cheek, that grant funding agencies should only pay for half the sequencing requested. The cost stopped falling earlier this year and actually went up when Illumina increased their pricing.
If you take 2500 into the mix as well then per base sequencing costs look like they have jumped by almost 15%. Personally I see 2500 as another step on the road to $1000 genomes and think the real price continues to fall but I'll say no more until the end of this post and my comment probably deserves a proper post all of its own some other time.
Genome Biology genomicist smack-down: Neil Hall's article After the gold rush in Genome Biology really nails the past five to ten years of genomics research, he specifically notes that he's being provocative and a little hard on us (he's a confirmed genomicist himself) before stating "We... have been spoiled. We have been real-estate agents working in a housing boom; bankers trading in debt. We have not been made to work; worse still, there has been very little incentive to think." Ouch! But he has a point. I've seen colleagues move on from the Institute where I work to good Universities and facing up to the realities of having to 'think' very hard when resources are more limited.
Almost anyone can discover something given unlimited resources. It would be interesting to attach a £ or $ sign to every research article and then measure output in more clearly economic terms. Which high-impact papers were the best value for money?
How cost-effective is "Collaborazilla" and his/her like?
Mick Watson on his blog goes over, under and reinterprets the NHGRI graph. He plots a new graph showing how price has changed when comparing time-points on the NHGRI graph. On it he notes that apart from the introduction of the GA, at almost every other time-point there has been only a modest drop in costs, and that over time the graph shows an upward trend.
The history behind the numbers: The graph everyone points to, and which Mick reinterprets, comes from the NHGRI and it shows a pretty steep fall from mid-2007 till early-2010. These were the "Solexa-years" from their 1G, through the Illumina acquisition and GAI, GAII, GAIIx (bananas and iPAR anyone) to HiSeq. The drop in costs were brought about by real technological improvements and as a user all the way through this drop it is easy to remember the upgrades, for all the good; as well as the pain some of them inflicted on my lab!
But the drop slowed dramatically in 2010 when HiSeq came out. Genome centres traded their entire GAIIx stock and suddenly had over-supply of capacity. The new instruments spat out billions of bases and initially users had problems filling all the available lanes. Then the "big-science" projects really got started and the data has flooded out ever since. If it weren't for $1000 genome noises by Life Technologies I doubt Illumina would have given us the 600G upgrades quite as quickly, we'd more likely have seen an Apple-esque dribbling out of technology gains over the past two years.
So what does CNN say: This is a news article for a news organisation and is aimed at the general public. The author Eilene Zimmerman at CNN Money is right in pointing out how much things have changed since James Watson's genome was sequenced in 2007 for an estimated $1million, and today you can sequence (but not analyse or interpret) a genome (30x) for about $3000 or $4000. But her commentary and interviewees give an all too rosy picture and I'm not sure anyone actually looked at the NHGRI graph judging by the comment "since 2007, the cost of genome sequencing has been in free-fall". Free-fall to me means a sustained downward trajectory, the NHGRI graph is perhaps showing the moment we reach terminal velocity.
The article also covers the impact of the Ion-Torrent technology and quotes Jonathan Rothberg saying "In three months, we'll be able to do one entire human genome for $1,000", I don't know of a single PGM or Proton customer who thinks this is close to reality. PGM has effectively stopped developing past the 318 chip and the Proton is a long-way from $1000 genomes. And of course ONT gets a look in as well although Eric Topol notes they have "significant problems with accuracy". Have we found an early-access customer perhaps?
But I agree with the articles hopes for genetic testing by sequencing and that the costs of these should drop to the level where health-care providers can roll them out nationwide. The costs of NGS are still coming down and are likely to drop for longer. I guess we need to spend more of our time and money on making sure we can execute with NGS tests in the clinic.
So what does the future hold? I'll be honest with an "I don't know" to start with. I am one of those bloggers that Neil refers to, excited by the prospect of new sequencing technologies and disappointed that ONT have run into difficulties. But the future is here in all its black and white glory. I'd say its HiSeq 2500.
Disclaimer: I run a lab that uses a lot of Illumina technology, but I don't have a 2500, and I'm not expecting Illumina will give me one however much I say it is the future!
But I do think HiSeq 2500 is the way forward, or at least part of the equation, because the costs of sequencing on this instrument could be lower in real-terms than anything we've seen so far. With the new patterned flowcells only six months away data volumes are going to jump up again. And the fact that HiSeq 2500 rapid run mode generates more data than standard mode in a given time-frame means the amortisation of capital purchase and service contract costs drop and these are a huge chuck of our real costs. Oh, and let's not forget the possibility of 1 billion 1000bp reads!
PS: Neil acknowledged comments or tweets as one of the reasons for his commentary article. I'd do the same and say this has been a topic for discussion since the middle of last year when genomicists get together. I'm guessing we'll still be talking about it for another year or two.