How to illustrate rise of sequencing productivity

[DebMil]

I'm new, so forgive me if this is not the right section to post this question (and I'd welcome suggestions for a better forum).

How can I illustrate how productivity of sequencing in a lab or the community has grown? I've seen dozens of talks that show the latest version of the same NHGRI slide of falling price of sequencing. Has anyone ever published or presented a description or graph of how many bases some core lab or genome center used to produce per day/week/month and how that has increased?

[GenoMax]

I am sure some labs have kept track of this over the years but they may not be comfortable publicly sharing this type of information. It may also be difficult to normalize this data (into unit productivity). Sequencers change output over time, new machines are added, sequencers may not run 24x7 (unless entity is a genome center).

[DebMil]

I guess I'm just trying to provide a sense of scale. The fall in price per base (or price per genome) does not really represent any single lab. It's a rough scale, but it's dramatic.

I was hoping for something like "core lab or genome center X used to have 15 ABI 3700s and could produce Y bases per day. Now the have 3 MiSeq's, a PacBio Model-A and and IonTorrent Model-B and with many fewer people they now generate Z bases per day. I want to give some anecdotal evidence to get across the idea that it's not only much cheaper now, but the work gets done much, much faster.

[SNPsaurus]

I think an interesting comparison would be to compare a top of the line sequencing center when the human genome was sequenced to now.
From https://www.celera.com/celera/pr_1056581295
Celera sequenced the human genome with 300 ABI Prism 3700 sequencers (according to http://en.wikipedia.org/wiki/Applied_Biosystems they cost $300,000 each). It looked like throughput was 10 plates per day and 500 bp per run, so throughput of one machine was 96wells x 10plates x 500bp = 480,000 bp per day x 300 machines = 144 million bps per day for a $90M investment in equipment.

A lab may have a NextSeq 500 ($250k), which can easily give 400M reads of 150 bp = 60 billion basepairs per day.

I haven't delved into the reagent costs for each of those runs, just the initial equipment costs. But a small machine, the kind that can be purchased by a single busy lab or a small core, easily is hundreds of times more powerful than the vast Celera warehouse used to produce the Celera human genome.

[GenoMax]

I would not bet on the fewer people part. Unless a lab is using significant level of automation (which would only be practical for large volume of samples) number of people needed (to QC samples/Make libraries/Run instruments/Manage data) is probably more than what a traditional sanger core needed in the past.

[DebMil]

Thanks, SNPsaurus. And GenoMax, I have to think that it took more people to run and supply Celera's 300 ABI 3700's than the one NexSeq 500, though I don't know what the staff levels had to be for either.

[Brian Bushnell]

Anecdotally, JGI used to have scores of people in a factory-like setting running a massive number of Sanger machines in 3 shifts, 24x7. Now we produce thousands of times as much data with less than half as many people directly involved with making libraries and running sequencing machines, during a normal ~40-hour work week (using Illumina and PacBio machines). There are probably more people analyzing the data now, though, since there's more data to analyze.

[GenoMax]

I was referring to a "core" lab and not a "genome center" where everything happens on a different scale.

This kind of comparison feels odd. Like saying that a current phone has a processor that is X% faster than an Intel 386DX CPU while being Y% more energy efficient.

I have been doing this too long

[SNPsaurus]

It is true that a core lab could make do with a single person running an ABI sequencer, whereas an NGS core might have several people (prep, sequencing, informatics).

And my comparison of Celera to a NextSeq is a bit apples to oranges. But I think it is pretty incredible that even the peak of power of the ABI era pales to a small machine now. For a small core, just take the output and divide by 300 for one machine and maybe 1/5th the sample throughput as well, so 1/1500 of Celera.

[westerman]

And yet ... we still have 3730s running every day. Power isn't everything.

[SNPsaurus]

Sure, no knock against ABIs... they just have a different role now than when a warehouse full of them represented the best way to sequence a genome.

[thermophile]

Is the goal of your core lab really maximum bases? I run a mix of amplicons and small genomes so bases sequenced doesn't really matter much. I do keep track of how many labs and departments we support as well as an ever growing list of applications that different labs ask us to help with. I feel that the bases per day metric makes more sense for a genome center (massive sequencing and only sequencing) rather than a core lab (developing methods/prelim data)