A few months ago, the National Human Genome Research Institute (NHGRI) updated their analysis of the cost of sequencing and, for the first time since records began, it got more expensive (Figure (Figure1).1). You know the graph, the one which looks like the profile of an aqua-park waterslide, a gradual incline followed by a precipitous drop as next generation sequencing kicks in. Well, now the waterslide ends with a treacherous upward flick! We have become so comfortable in the knowledge that DNA sequencing reduces in cost at a rate that makes each run cheaper than the last, that some of the scientific community are in denial. I have even seen people present this graph at meetings and explain how sequencing is getting cheaper every day despite the fact they are standing in front of a 10 foot PowerPoint slide showing clearly that this is not true. In fact, the cost of sequencing a human genome increased by $717 (an increase of 12%) between April 2012 and October 2012. This month the new figures showed that the price fell again, but the point remains—you can forget Moore’s law! Some of you will think this merely means you need to replace the opening slide in your PowerPoint deck and tone down some of the rhetoric around $10 human genomes and the advent of free sequencing. I, however, think that the long-term ramifications may be more profound...‘But...‘, I hear you scream, ‘...this is a temporary blip. Soon we will be saved by new cool technology that will plug into my laptop and sequence a genome for $10 in an hour’. In reality, is this just something we simply want to believe? There really is no reason to think that sequencing methodology is about to undergo a revolution in the near future. I am always amazed at the self-inflicted hype that follows any hint of a story where some company has come across a new way of sequencing that is going to turn all our Illumina kits into oversized doorstops. Often this comes not from the companies themselves but the scientists who are so desperate to buy them. The hype is usually followed by hyper-critical twitter and blog commentaries when the machine in question does not appear to do what we want it to (see this revealing interview with Oxford Nanopore’s Clive Brown), in a cycle that has repeated itself at least three times in the last 5 years. I begin to wonder why we don’t learn from history.
This graph may or may not tell a different story. The story is that yes, sequencing costs are coming down; but since late 2007, early 2008 the rate of change of that reduction has been following an upwards trend i.e. over time, the reduction in cost from one period to the next has been increasing.
I’m going to try and lay this out in a completely technology neutral way, though I will have to mention different sequencing technologies at some point. However, I am pretty convinced of this one fact: there is not a single sequencing technology out today that can deliver 30X of a human genome for anywhere near $1000....None of the current sequencing companies can deliver 30x of a human genome for less than $1000 reagent costs (using list prices) Yes, that’s right – even ignoring points 2-5, even just buying reagents, the cost is greater than $1000 for a 30x human genome. Now, it’s possible Broad, BGI, Sanger etc can get below $1000 for the reagents due to sheer economies of scale and special deals they have with sequencing companies – but then remember they have to add in those extra charges (2-5) above. Obviously, Illumina don’t charge themselves list price for reagents, and nor do LifeTech, so it’s possible that they themselves can sequence 30x human genomes and just pay whatever it costs to make the reagents and build the machines; but this is not reality and it’s not really how sequencing is done today.
In my recent talk at the NIH symposium to mark the 10th anniversary of the HGP: http://bit.ly/KDHGP10 … I quoted a personal communication from Illumina CSO David Bentley, who says that in batch mode, the HiSeq can currently sequence five human genomes (presumably to 30x or higher) for a reagents list price of $25,000 — or $5,000/genome. With negotiated discounts (or if you want to estimate the wholesale cost), take 1/3 or 1/2 of that figure. So for what it’s worth, we might be edging close to the $2,500 genome, but that’s as good as it gets for now.
“He’s right,” Topol said. “If you get a bunch of genomes done at Illumina, you can get ’em for $2,500 each—today.” But in 2004, Topol said, sequencing a human genome cost $28.7 million. “Now we already have a 99.8 percent plus price reduction,” Topol said. “We don’t have that much further to go to get from $2,500 to $1,000. Most everyone would forecast that in a couple of years we will get to that number, with deep coverage of 40-fold, so it’s accurate. I think it’s clearly within reach now.” “And I want to take it a step further,” Topol said. “It’s going to go well below $1,000 a genome in the future.”...Topol agreed with Watson’s take on the PeerJ statement [$100]. “That’s a little far-fetched,” he said. But incremental progress in getting the price lower once that $1,000 mark has been reached will continue.
$100 might be a little questionable here (apparently Illumina has a history of making the most favorable possible assumptions about volume/amortization) but revisiting my original prediction from 7 years ago:
As part of my work for Luke, I looked into price projections for whole genome sequencing, as in not SNP genotyping, which I expect to pass the $100 mark by 2014. The summary is that I am confident whole-genome sequencing will be <$1000 by 2020, and slightly skeptical <$100 by 2020.
I was too pessimistic about SNP genotyping (it was actually more like $50 in 2014, I was completely unaware of UK Biobank at the time or its scale or savings), definitely right about ‘<$1000 by 2020’, and I think I will turn out to be somewhat wrong about WGS being <$100 by 2020: even if Illumina is fudging some numbers for early 2019 at $100, it’ll have almost a whole year to drop the cost a little more, and honestly, even if it’s actually $110 does it make a difference considering how many things you can use whole genomes for & general medical overhead? You can hardly get some prescription aspirin these days for $100...
Overall self-assessment: I was more right than I had any right to be in that set of predictions given I was using some simple extrapolating and adding some pessimism/mean-reversion. Not bad, past-self!
Veritas Genetics will be offering their MyGenome product (30x whole genome sequencing) normally $999, for $199 to the first 1000 customers, starting tomorrow, Monday, 9 AM ET.
Even allowing for promotional discounts, I’m still impressed. EDIT: Dante Labs too!
Thanks! I’ve been conflicted about which SNP service to use, and now I don’t have to decide. :) Do you know if there are any potential downsides to consenting to let Veritas use the data for research? Would you tick that box?
I am not sure you necessarily want to use Veritas/Dante Labs (Veritas might be sold out already based on their Twitter), as WGS reports are usually pretty raw and you won’t get all of the interpretive services somebody like 23andMe would provide. I don’t believe 23andMe or the other major services let you just upload sequencing data either, only download. Offhand, I’m not sure how easy it would be to even use Promethease (not that Promethease is very worthwhile, as most of their report is candidate-gene junk). Personally, I am holding off on getting a WGSes done. I don’t know what I would do with mine, and the price should keep getting lower.
Oh, I misunderstood the purpose of your comment and thought you were recommending people to take advantage of the sale. I knew it was going to sell out quickly so I made the order prior to posting my question. (I gave consent for research since it said that I could withdraw that consent at any time.)
It looks like Veritas offers VCF file download so it’s compatible with Promethease but the format it uses only gives 16,000 genotypes. Also apparently Veritas used to provide the full BAM raw data, but no longer does, which is disappointing, so I’ll probably cancel my order and take advantage of the Dante $199 sale instead which does offer BAM. Looks like sequencing.com lets you upload a BAM file and offers a bunch of apps to do different analyses on it.
No, I was mentioning the sales because they offer a measurement of what WGS costs end to end now—presumably Veritas/Dante or Nebula are offering at close to their marginal cost (as they aren’t big or wealthy enough to afford to give it away and WGSes aren’t exactly a repeat-customer business). As far as Dante goes, I have seen some complaints about very slow or inconsistent service; on IRC, one of us did a previous sale and their original spit didn’t work, so they sent him another tube and forgot the postage. Not sure if he’s gotten his WGS yet either.
I see. Given that I haven’t done a genotype yet, would you suggest that I go through with Dante anyway, or wait until the price comes down further? (Presumably it would definitely be worth doing at $100?)
Well, do you have anything in mind specifically to do with it? If you do, it may not be worthwhile to wait. But if you don’t have something which needs to be done with a WGS right now, you probably aren’t going to be struck with inspiration once you get your download either.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3663089/
http://biomickwatson.wordpress.com/2013/05/15/a-pedantic-look-at-the-cost-of-sequencing/
http://biomickwatson.wordpress.com/2013/06/18/the-1000-myth/
http://biomickwatson.wordpress.com/2013/06/18/the-1000-myth/#comment-2031
http://www.utsandiego.com/news/2013/Jun/19/1000-genome-mirage/2/?#article-copy
Consumer WGSes hit ~$1000 with Veritas in 2016. In 2018, Dante Labs began offering WGS at ~$600, with a sale of $350. And we now have a rumor that Illumina will announce a $100 genome in a few months (presumably in early 2019): https://twitter.com/coregenomics/status/1058790189752049664
$100 might be a little questionable here (apparently Illumina has a history of making the most favorable possible assumptions about volume/amortization) but revisiting my original prediction from 7 years ago:
I was too pessimistic about SNP genotyping (it was actually more like $50 in 2014, I was completely unaware of UK Biobank at the time or its scale or savings), definitely right about ‘<$1000 by 2020’, and I think I will turn out to be somewhat wrong about WGS being <$100 by 2020: even if Illumina is fudging some numbers for early 2019 at $100, it’ll have almost a whole year to drop the cost a little more, and honestly, even if it’s actually $110 does it make a difference considering how many things you can use whole genomes for & general medical overhead? You can hardly get some prescription aspirin these days for $100...
Overall self-assessment: I was more right than I had any right to be in that set of predictions given I was using some simple extrapolating and adding some pessimism/mean-reversion. Not bad, past-self!
Just got a Veritas-related email:
Even allowing for promotional discounts, I’m still impressed. EDIT: Dante Labs too!
Thanks! I’ve been conflicted about which SNP service to use, and now I don’t have to decide. :) Do you know if there are any potential downsides to consenting to let Veritas use the data for research? Would you tick that box?
Yes. In fact, I am already a PGP participant.
I am not sure you necessarily want to use Veritas/Dante Labs (Veritas might be sold out already based on their Twitter), as WGS reports are usually pretty raw and you won’t get all of the interpretive services somebody like 23andMe would provide. I don’t believe 23andMe or the other major services let you just upload sequencing data either, only download. Offhand, I’m not sure how easy it would be to even use Promethease (not that Promethease is very worthwhile, as most of their report is candidate-gene junk). Personally, I am holding off on getting a WGSes done. I don’t know what I would do with mine, and the price should keep getting lower.
Oh, I misunderstood the purpose of your comment and thought you were recommending people to take advantage of the sale. I knew it was going to sell out quickly so I made the order prior to posting my question. (I gave consent for research since it said that I could withdraw that consent at any time.)
It looks like Veritas offers VCF file download so it’s compatible with Promethease but the format it uses only gives 16,000 genotypes. Also apparently Veritas used to provide the full BAM raw data, but no longer does, which is disappointing, so I’ll probably cancel my order and take advantage of the Dante $199 sale instead which does offer BAM. Looks like sequencing.com lets you upload a BAM file and offers a bunch of apps to do different analyses on it.
No, I was mentioning the sales because they offer a measurement of what WGS costs end to end now—presumably Veritas/Dante or Nebula are offering at close to their marginal cost (as they aren’t big or wealthy enough to afford to give it away and WGSes aren’t exactly a repeat-customer business). As far as Dante goes, I have seen some complaints about very slow or inconsistent service; on IRC, one of us did a previous sale and their original spit didn’t work, so they sent him another tube and forgot the postage. Not sure if he’s gotten his WGS yet either.
I see. Given that I haven’t done a genotype yet, would you suggest that I go through with Dante anyway, or wait until the price comes down further? (Presumably it would definitely be worth doing at $100?)
Well, do you have anything in mind specifically to do with it? If you do, it may not be worthwhile to wait. But if you don’t have something which needs to be done with a WGS right now, you probably aren’t going to be struck with inspiration once you get your download either.