MiSeq

[james hadfield]

The latest addition to the Ilumina stable! What's next?

Illumina have just released the MiSeq instrument which allows a very quick turnaround on runs that may be useful for small genomes, QC runs of larger projects and personalised medicine.

I could not see anything about reagents though and wonder if this is a single lane system or not. It is only going to image about 10% of a lane based on what we see in GAIIx or HiSeq. Of course the lane might only be 10% of the size of a HiSeq flowcell.

There is also little about how it achieves such speed. The only thing they say in teh documentation is that new fluidics help speed things up, I am not sure how this makes an 8 fold difference though.

1 x 35bp 4 hours >120 Mb
2 x 100bp 19 hours >680 Mb
2 x 150bp 27 hours >1 Gb

Who's first?

[krobison]

My personal guess is that the smaller flowcell surface simply enables faster scanning. Anyone know how long it takes a HiSeq to scan the flowcell?

Cost per run is claimed to be in $400-$750 range, which sure looks nice. 6.8M reads per run.

[james hadfield]

I know how long it takes a GAIIX to scan a flowcell, although I am not completely sure about the HiSeq. It is chemistry time that is most important here as this is more difficult to change. We have done a PE150bp run on a subset of tiles where we did not need such a volume of data and run times dropped significantly.

First some 'facts':
1) The cycle time is about one hour on GAIIx and is faster on HiSeq; this includes chemistry and imaging.

2) Imaging takes just under three seconds per tile on GAIIx, there are no 'tiles' on HiSeq. This would equate to about 45 minutes for imaging and 15-20 minutes for chemistry.

3) If MiSeq is only giving 1Gb from a PE150bp rune then it probably has about 10% the capacity of a GAIIx or HiSeq flowcell.

Now my 'personal guess':
MiSeq would require a significant change to the fluidics to bring chemistry cycling times down as low as needed to get 2x150bp or 300 cycles in 27 hours. This only allows 5m40s for each cycle. But this might be achievable on a miniature flowcell with better, more direct Fluidics. The imaging for ten ‘tiles’ would be about 30sec.

This certainly means little hope of bringing the time for HiSeq down as much unless some of the improvements are backwardly compatible.

I guess we need to wait for someone outside of Illumina to run one. I would be willing of course if Illumina need an early access tester!

[nickloman]

They do say it is the fluidics that have enabled the faster runs

http://www.genomeweb.com/sequencing/illuminas-low-cost-miseq-promises-speed-next-gen-sequencing

A few extra thoughts on MiSeq from my blog:

http://pathogenomics.bham.ac.uk/blog/2011/01/miseq-now-thats-what-im-talking-about/

We do indeed live in interesting times ...

[kmcarr]

3) If MiSeq is only giving 1Gb from a PE150bp rune then it probably has about 10% the capacity of a GAIIx or HiSeq flowcell.

It's much worse than that. According to the data sheets for each instrument (and my GAIIx and HiSeq might be out of date):

HiSeq2000: 1 billion PF clusters per run
GAIIx: 320 million PF clusters per run
MiSeq: 3.4 million PF clusters per run

So the MiSeq is ~ 1% the capacity of a GAIIx and ~0.33% the capacity of a HiSeq2000

[konrad98]

kmcarr - I think James was referring to lanes. A PE lane on a GAII will yield around 70-80 million reads. If the MiSeq yields 7 million per run then the numbers look fine.

[kmcarr]

kmcarr - I think James was referring to lanes. A PE lane on a GAII will yield around 70-80 million reads. If the MiSeq yields 7 million per run then the numbers look fine.

See James' quote; he specifically relates the capacity to a GAIIx or HiSeq flowcell.

[csc]

James, I can confirm that it is a single lane system.

[GW_OK]

See James' quote; he specifically relates the capacity to a GAIIx or HiSeq flowcell.

Let's not forget the upcoming AGBT meeting. I wouldn't be surprised if Illumina ups the HiSeq Readlength to 150bp and/or increases cluster density.

[HESmith]

My money is on increased cluster density. The current recommended maximum for the HiSe2000 is 425K/mm2, vs. 700-800K/mm2 for the GAIIx. Plus, the chemistry improvements have not kept pace with software improvements in the past, and the chemistry is still the limiting factor for longer runs.

But I hope I'm wrong...

[ECO]

SEQanswers volunteers to put a MiSeq in its garage to sequence community samples!

[csc]

I also think it is interesting that the MiSeq is apparently the progeny of the Avantome acquisition. I'm not sure what specific tech. they used from Avantome IP, perhaps the fluidics? I'm wondering if Illumina still plans on some sort of long read platform based on Avantome IP as they said when they first made the acquisition.

[csc]

Well, not surprisingly, it looks like the most immediate improvement is going to be increased cluster density. Flatley says upgrades (software/reagents) this spring will give HiSeq 2000 throughput of 600 gigabases. Software is not surprising, they are probably just pushing cluster densities to GAII levels. Oddly enough, the new sequencing chemistry will initially only support 2x100bp instead of 2x150bp. Internally Illumina has achieved 1.13 terabase runs using 2x150bp reads over 14 days. This was all reported at the JP Morgan Healthcare conference out here in CA. I wonder what kind of exciting stuff they have to talk about at AGBT...

[system7]

MiSeq Run Times

Hi All,

I thought I would clarify what is included in the MiSeq run times:

1x35 - 4hrs. This time includes cluster gen and sbs
2x100 - 19 hrs. This time includes cluster gen and sbs and PE turn
2x150 - 27 hrs. This time includes cluster gen and sbs and PE turn.

In terms of how fast the system runs, the chemistry cycle time has been reduced 5x. Imaging a smaller area relative to GA or HiSeq is only half the battle, it is the chemistry time reduction that is the other half of the equation in yielding such a fast system.

Rob

[james hadfield]

Anyone want to bet on how high Illumina are going to predict HiSeq will go by the end of 2011?

My pick in the sweepstake is 2Tb, it is HiSeq 2000 after all!