Why Has The Cost Of Genome Sequencing Decline So Rapidly?

Question

[Khader Shameer]

What was the primary experimental, technical or computational factors that led to the rapid decline of genome sequencing cost ?

[Anima Mundi]

Hi, you may want to have a look at this: http://www.genome.gov/sequencingcosts/

[Istvan Albert]

I saw this graph before, now that I looked more closely it seems a bit overly optimistic, especially the right hand side - cost per genome.

[Anima Mundi]

I was thinking about founding a company :P.

[Daniel Standage]

Equally interesting and informative would be the related economic and social factors.

Answer 1

[Jeremy Leipzig]

I suppose the pivotal moment was the bold decision of ABI (then PE Biosystems) to sequence the human genome on their own using capillary sequencing machines, turning the Human Genome Project into a race. As I understand it this was almost entirely due to the efforts of Hunkapiller and Venter. The Human Genome Project really floundered until this moment. We would still be waiting for a first draft of the human genome today if competition was not a factor, and there would be no incentive to develop the new sequencers that followed. So I would attribute the advances to ego.

[Philippe]

I don't totally agree. Even though that might be true for some period, the biggest drop in sequencing costs is essentially due to the emergence of "next generation" high-throughput sequencing technologies. Previous efforts (during the Human Genome Project for example) were more oriented to make Sanger sequencing more efficient (automatisation for example), the throughput per sequencing reaction did not improve much (relatively to the last "revolution"). The emergence of these new technologies is also a (economical) competition that stimulated the innovation.

[Jeremy Leipzig]

In terms of NGS, I would attribute the advances, beginning with MPSS, primarily to imaging technology. Fluorescence had been mastered for some time before dyed nucleotides, but I don't think digital cameras had the resolution to see beads until around 2000. I still think the genesis of all this was that board meeting at PE where they said "let's just do it".

[Khader Shameer]

Nice thought Jeremy. Is MPSS = Massively Parallel Sequence by Synthesis ?

[Jeremy Leipzig]

http://en.wikipedia.org/wiki/Massively_parallel_signature_sequencing

[Larry_Parnell]

I agree with Jeremy on this. I was present at the CSHL Biology of Genomes meeting at the time Venter made his announcement. The shock was deeply felt and did indeed spur the publicly funded project to action. The question asks about the primary experimental, technical or computational factors, but sometimes other forces come into play to put those three into practice and actually be used. The race was one such factor.

[Philippe]

I agree with Jeremy and Larry that the initial drive to have more efficient and cheaper sequencing technologies was the HGP and the competition of Craig Venter (glad to hear some insights from people that were actually there). But I still think that, even though they are associated of this willingness to rapidly sequence DNA, the development of some technologies was independent from that event. For example (but I might be wrong, thanks to correct me if so), I don't think (or don't know?) that technologies based on the detection of electric signatures of nucleotide incorporation was planned at the time.

Answer 2

[Philippe]

Hi,

to my opinion here are some factors that contributed to decrease sequencing costs (some were already mentioned). Note that not all points can be applied to all technologies.

• Parallelization of sequencing reactions. Technologies such as Illumina, IonTorrent or Nanopore use technologies that can easily be parallelized (clusters of oligos, microchip and nanopores, respectively). This allows to run a large number of parallel sequencing reactions in a relatively small instrument. With Sanger sequencing, to parallelize reactions you need a huge amount of individual sequencers.
• Less expensive chemistry. Most of new technologies (e.g. IonTorrent, Nanopore, 454) sequencing libraries do not rely on modified nucleotides (ddNTPs) that are actually expensive. I don't have any numbers to provide you with, but I guess this can be found online. This also (sometimes) implies shorter protocols for libraries preparation which therefore decrease the human cost. Some technologies such as Oxford Nanopore Technologies (platforms are not commercialized yet) do not even need some expensive chemistry since they are mostly relying on the electric signature of nucleotide interacting with the nanopore. In general, excluding the presence of ddNTPs and/or dNTPs considerably reduces the cost of sequencing.
• Increase in throughput. The decrease in sequencing cost is tightly linked to the increase in sequencing throughput of the "next generation" sequencers. In one single run (that still can take days or weeks) one machine can sequence several gigabases of DNA while previous technologies (i.e. Sanger-based methods) were more in the order on megabases.
• Absence of fluorescent signal detection. Sanger, Illumina or 454 technologies need to detect fluorophore signal to know which nucleotide was inserted during the sequencing reaction. Lasers and scanners that fulfill the accuracy requirements are very expensive. That, in part, explains the discrepancy of prices between an Illumina (fluorophore-based reaction) machine (~500K$) and plaforms such as IonTorrent (from ~50k$ per machine- microchips detect nucleotide incorporated) or ONT platforms (announced price of ~30k$ per cluster - nucleotide detection also rely on electric signal detection).

Some data come from this comparative table of sequencing technologies

Disclaimer: I didn't mention technologies such as Pacific Biosciences or Helicos since I am not familiar/documented with these technologies.

[Khader Shameer]

Thanks Philippe, this is a very nice summary.

Answer 3

[Raygozak]

I would say there are a lot of factors that contributed into the declining of the costs, since we are talking of sequencing this would be mostly in the experimental and chemical advancements, and also with the engineering part of things. Computational advances might have helped, but i would define sequencing only on the basis of defining molecular/physical strategies to make the nucleotide identification. It would be hard to quantify each one since they vary across platforms and methodologies. Another factor which is outside of the science realm is the law of offer and demand since the fact that nowadays there are more companies working on this, there is tight competition to bring the next thing in sequencing that scientist would find appealing and invest their money in.

[Raygozak]

I think it would be really hard to come up with precise items to describe this, since companies would really tell you about the same vague answers, since specific reasons would be trade secrets.

[Khader Shameer]

"... engineering part of things" I am really curious to know more about those "things".

[Larry_Parnell]

In this regard, modification and improvement of the dyes used to recognize and distinguish the four (or five if you allow for RNA synthesis with uracil) is important from an engineering side and from the trade secret side.

[Khader Shameer]

Larry, I would really really love to hear your perspective on this as a separate answer based on your experience.

[Raygozak]

I'm sorry i'm referring to the machinery used in creating the sequencing machines, there might be better materials developed and better techniques to multiplex the sequencing, or microfluidics techniques, to name a few.

Answer 4

[Mary]

I would say automation. The robotics and workflows to just keep pumping new input into the system. Obviously each subsequent increase in performance of sequencing tech added on--and will keep doing so. But primarily--I'd say automation.

I wonder how many people here ever loaded a sequencing gel...and sat in the back of the lab with a colleague reading off the bases. Good times. And a giant time sink.

[Larry_Parnell]

Exactly why E. coli was not the first bacterial genome to be completed. Fred Blattner and his group were running these huge (about 3-feet, 1-meter long) acrylamide gels while other genomes started later and finished ahead of his.