[MrGuy]

Phew. Thanks for bringing me back to Earth. I felt like I was on Mars for a while there.

With: "when I calculate 660M*.75*200 I get closer to 100GB"
Then we get 200GB with a 400bp chemistry. Ok, this feels much more realistic.

[scbaker]

75% active?

I haven't looked at this very hard but it does seem like your math is wrong. You say that a 660M feature chip at 200 bp per feature at 75% is 1000GB but when I calculate 660M*.75*200 I get closer to 100GB. All your numbers seem to be off by a factor of 10.

100GB is about 30x human genome coverage. Which is what I would consider the minimum for good variant coverage. So I would be willing to say closer to 2 human genomes per machine per day. 600 per machine per year. BGIing it at 138 machines, say 82,000 genomes per year.

Feel better?

These numbers look right (and are in line with the table ECO posted originally). My question is how many people are getting 75% active beads on the PGM chips? With 200b reads, this would yield the following:

314 = 0.9M*200b=180Mb (original spec = 10Mb)
316 = 4.6M*200b=920Mb (original spec = 100Mb)
318 = 8.3M*200b=1.7Gb (original spec = 1Gb)

I haven't seen people talking about 314 and 316 outputs that high. Is anyone here getting these kind of outputs?

[Elcannibal]

A little more math to really mess with you: Ion is saying they can do the genome for $1000. However, is this with the calculation of 200 bp or 400 bp at the less efficient live wells rate? Let's assume 75% live wells... That means the human genome could be done 10-20 at a time on the 660M chip for $50-$100 per genome. And the proton does a run 2 times a day. Max 40 genomes per day. 300 working days per year. That's 12,000 genomes per machine. If BGI bought 138 of these, that would be 1,656,000 genomes per year.

Please, someone check my math. I feel ill.

A genome could be free to sequence, it doesn't change much of the larger challenges out there, let alone the fact that no public company will survive from a reverse razor and blade system...

[nickloman]

These numbers look right (and are in line with the table ECO posted originally). My question is how many people are getting 75% active beads on the PGM chips? With 200b reads, this would yield the following:

314 = 0.9M*200b=180Mb (original spec = 10Mb)
316 = 4.6M*200b=920Mb (original spec = 100Mb)
318 = 8.3M*200b=1.7Gb (original spec = 1Gb)

I haven't seen people talking about 314 and 316 outputs that high. Is anyone here getting these kind of outputs?

Sorry Shawn, I should clarify. We're getting *loading* of about 75-80%, about 2/3 of these are live, and about half in total pass quality filters. So far we've been using 100bp reads. So we're looking at 3m / ~300Mb on 316. I know some groups have got as high as 500Mb.

Here are results from our first 316 runs:

http://pathogenomics.bham.ac.uk/blog/2011/07/ion-torrent-316-first-impressions/

[scbaker]

Sorry Shawn, I should clarify. We're getting *loading* of about 75-80%, about 2/3 of these are live, and about half in total pass quality filters. So far we've been using 100bp reads. So we're looking at 3m / ~300Mb on 316. I know some groups have got as high as 500Mb.

Here are results from our first 316 runs:

http://pathogenomics.bham.ac.uk/blog...t-impressions/

Thanks, Nick. That clarification is very helpful!

[epistatic]

I shouldn't plan for any NGS instrument to last longer than 1 year, maybe 2? I manage a NGS Core facility and looking at pricing to possibly purchase an Ion Proton.

For our first year, the proton + one touch + server = $250K.
If I forget depreciation and say this instrument is only good for 1 year and have to recover the "true cost" of sequencing, the $1K genome is not yet here. Ignoring labor/consumable costs and only factoring in $1K all inclusive genome run (prep-seq) and the instrument purchase price, I'd safely say we are still in the $2K - $5K genome with the Proton.

For 2012 there are 252 work days in the year.

If I run proton 100% = $992 instrument cost recovery + $1K run = $1992/genome
If I run proton 75% = $1323 instrument cost recovery + $1K run = $2323/genome
If I run proton 50% = $1984 instrument cost recovery + $1K run = $2984/genome
If I run proton 25% = $3968 instrument cost recovery + $1K run = $4968/genome

Of course the price goes down if I also factor in instrument recovery over a second year, but also have a service contract and handful of other possible competing NGS instruments out there to replace it. The PGM lasted <1 year, is Proton II already in R&D? If sequencers come out like iPhones each year we are all in trouble! The PGM was only $50K, but how much did that add to each run price?

I think it still bodes well for sequencing cores and service centers. The Proton isn't cheap enough for single labs to purchase and run if they get passed the marketing spin and think about all inclusive run costs. It is still more cost effective to send your samples out and let someone else worry about hot new instantly replaced expensive sequencing tech.

[scbaker]

I shouldn't plan for any NGS instrument to last longer than 1 year, maybe 2? I manage a NGS Core facility and looking at pricing to possibly purchase an Ion Proton.

For our first year, the proton + one touch + server = $250K.
If I forget depreciation and say this instrument is only good for 1 year and have to recover the "true cost" of sequencing, the $1K genome is not yet here. Ignoring labor/consumable costs and only factoring in $1K all inclusive genome run (prep-seq) and the instrument purchase price, I'd safely say we are still in the $2K - $5K genome with the Proton.

For 2012 there are 252 work days in the year.

If I run proton 100% = $992 instrument cost recovery + $1K run = $1992/genome
If I run proton 75% = $1323 instrument cost recovery + $1K run = $2323/genome
If I run proton 50% = $1984 instrument cost recovery + $1K run = $2984/genome
If I run proton 25% = $3968 instrument cost recovery + $1K run = $4968/genome

Of course the price goes down if I also factor in instrument recovery over a second year, but also have a service contract and handful of other possible competing NGS instruments out there to replace it. The PGM lasted <1 year, is Proton II already in R&D? If sequencers come out like iPhones each year we are all in trouble! The PGM was only $50K, but how much did that add to each run price?

I think it still bodes well for sequencing cores and service centers. The Proton isn't cheap enough for single labs to purchase and run if they get passed the marketing spin and think about all inclusive run costs. It is still more cost effective to send your samples out and let someone else worry about hot new instantly replaced expensive sequencing tech.

It looks like you're assuming one run per day. I've seen the run length being quoted as anything from 2 hours to "a few hours". Even if the run is 4 hours, it seems like you could fit in more than one run per day. That would help bring your amortization costs down.

[epistatic]

Two a day could be fair if the set up, take down, cleaning, etc. was quick. I wouldn't predict a 100% utilization even if it could do 10 runs in a week. If you could fragment DNA and load onto the chip for sequencing then you likely could keep up. It will be interesting to review the entire workflow and run specs. If the read length increases to 400 bp as predicted, the run times will have to scale some versus current 100 bp reads unless their fluidics and signal processing greatly increase in speed. Very cool product, I'm not sure how they can market a SOLiD now.

[Mirth]

Question, how do you fragment the library, are you using sonication or the IonShear. We are trying to buy a MiSeq/PGM but we can use Nextera with the MiSeq, with the PGM do you need a sonicator? I mean requiring a Covaris would certainly put the PGM out of our reach...

You can use Ion Shear, NEB for shotgun libraries. Ampliseq for target reseq and Ion RNA-seq kits.

[scbaker]

Does SOLiD have a future?

Two a day could be fair if the set up, take down, cleaning, etc. was quick. I wouldn't predict a 100% utilization even if it could do 10 runs in a week. If you could fragment DNA and load onto the chip for sequencing then you likely could keep up. It will be interesting to review the entire workflow and run specs. If the read length increases to 400 bp as predicted, the run times will have to scale some versus current 100 bp reads unless their fluidics and signal processing greatly increase in speed. Very cool product, I'm not sure how they can market a SOLiD now.

Based on Life's public comments, it appears that they've (mostly) given up on marketing the SOLiD. They did, however, mention launching a new prep system for SOLiD this spring which would eliminate the beads and increase the feature density ~5-fold. It will be interesting to see what comes of that.

[DICER]

Two a day could be fair if the set up, take down, cleaning, etc. was quick. I wouldn't predict a 100% utilization even if it could do 10 runs in a week. If you could fragment DNA and load onto the chip for sequencing then you likely could keep up. It will be interesting to review the entire workflow and run specs. If the read length increases to 400 bp as predicted, the run times will have to scale some versus current 100 bp reads unless their fluidics and signal processing greatly increase in speed. Very cool product, I'm not sure how they can market a SOLiD now.

SOLiD has different chemistry and error correction ... bing bing.

But for the Ion Proton, a decent grant application will see Ion Protons in a high number of labs, especially via the upgrade option

,:0)

[DICER]

Phew. Thanks for bringing me back to Earth. I felt like I was on Mars for a while there.

With: "when I calculate 660M*.75*200 I get closer to 100GB"
Then we get 200GB with a 400bp chemistry. Ok, this feels much more realistic.

on other thing to factor into the discussions is that they are all single end runs and the paired end protocol essentially doubles the output

[SeqAA]

Two a day could be fair if the set up, take down, cleaning, etc. was quick. I wouldn't predict a 100% utilization even if it could do 10 runs in a week. If you could fragment DNA and load onto the chip for sequencing then you likely could keep up. It will be interesting to review the entire workflow and run specs. If the read length increases to 400 bp as predicted, the run times will have to scale some versus current 100 bp reads unless their fluidics and signal processing greatly increase in speed. Very cool product, I'm not sure how they can market a SOLiD now.

The run times will probably scale a little, but it hasn't been linear. 100bp is down to 1.5 hrs in 2.0, and 200bp is not 3 hrs. Originally 100bp was just under 3 hours, so something is getting faster.

[Elcannibal]

Is there a Life Grand Challenge for the Proton...

Maybe a Billion $$$ to cure all human diseases ?

Joking aside, how does this change the infamous and untouchable Grand Challenges on the PGM (you know, it's hard to be serious on that one)...

[DICER]

that's why they call it a Grand Challenge ... bom bom ;0)