[scbaker]

Overall my conversation was pretty disappointing as far as new info (although the gentleman I talked to resisted a pretty good grilling from me and a 454/Roche product manager).

Points from my conversation that I can remember...

• GridION == "4-5x more sensors" than the MinION (no comparative runtime specs).
• GridION can sip from a 96 well plate full of samples (this was demo'd as a flat bottom black microtiter plate with no septa...hmmm).
• Acquisition speed == "bases per second".
• The above picture suggesting sample prep is just a "10bp overhang" is not correct and not confirmed
• The prototypes of the MinION ended up much larger than shown at AGBT due to heat issues (had to "add fans").
• They will not confirm which gating protein (exo or phi29) or which pore (mspA or a-HL), but that the gate protein is not covalently coupled to the pore. (which strongly suggests to me that it is phi29...the exo-release method would undoubtedly need exquisite positioning for efficient detection)
• According to the CEO, the DNA input concentration is "nanomolar". This makes sense to be given the binding kinetics of phi29 to it's binding site...but translates into high concentrations for a single molecule platform. When questioned about the possibility of doing very dilute samples where all the molecules matter, he suggested some method of on-chip enrichment/concentration (?????).
• They are cagey about data release because they want to (my paraphasing) "release data when it's ready"

Anyway, that's all I got. The ratio of ([booth girls] + [Apple-esque brushed aluminum] + [blinky lights]) to (data) was undefined...so that seems to speak for itself.

Thanks for summarizing for those of us who couldn't be at ASHG. All indications up to now have been for ONT using a highly modified version of a-HL, so I'd be surprised if the first version actually uses mspA. In terms of the exonuclease method, I'm pretty sure it's been abandoned (or put waaaay back on the backburner - hence the 'arbitration hearing' between ONT and Illumina).

[ECO]

Since data = 0, don't you mean infinite?

If it were approaching zero, it would be approaching infinity. It was zero...and x/0 == undef.

[HESmith]

If it were approaching zero, it would be approaching infinity. It was zero...and x/0 == undef.

Touche! (or, in the immortal words of Buzz Lightyear, "To inifinity and beyond!")

Definitely getting punchy...

[Nanoporous]

Overall my conversation was pretty disappointing as far as new info (although the gentleman I talked to resisted a pretty good grilling from me and a 454/Roche product manager).

Points from my conversation that I can remember...

• GridION == "4-5x more sensors" than the MinION (no comparative runtime specs).
• GridION can sip from a 96 well plate full of samples (this was demo'd as a flat bottom black microtiter plate with no septa...hmmm).
• Acquisition speed == "bases per second".
• The above picture suggesting sample prep is just a "10bp overhang" is not correct and not confirmed
• The prototypes of the MinION ended up much larger than shown at AGBT due to heat issues (had to "add fans").
• They will not confirm which gating protein (exo or phi29) or which pore (mspA or a-HL), but that the gate protein is not covalently coupled to the pore. (which strongly suggests to me that it is phi29...the exo-release method would undoubtedly need exquisite positioning for efficient detection)
• According to the CEO, the DNA input concentration is "nanomolar". This makes sense to be given the binding kinetics of phi29 to it's binding site...but translates into high concentrations for a single molecule platform. When questioned about the possibility of doing very dilute samples where all the molecules matter, he suggested some method of on-chip enrichment/concentration (?????).
• They are cagey about data release because they want to (my paraphasing) "release data when it's ready"

Anyway, that's all I got. The ratio of ([booth girls] + [Apple-esque brushed aluminum] + [blinky lights]) to (data) was undefined...so that seems to speak for itself.

Thanks for the very nice summary.

For the 'gate' protein, I think they are using some sort of a motor protein to slow down the motion of the DNA in a stepwise manner. As per Clive Brown's talk at AGBT, it is certainly not attached to the pore, and that it is definitely not phi29 (which is what the Akeson and Gundlach groups are using). A polymerase like that would require that you add NTPs. So probably a helicase or something.

[TonyBrooks]

Maybe ONT have learnt from PacBio's mistakes. Remember the anticlimax when you first saw PacBio data?

Nanopore error rate is around 4% ~Q14, or at least it was 9 months ago.
The problem is we've all been spoilt by Illumina SBS and will see anything with < Q30 (or even Q20). They've gone on record saying they won't release anything until error is <1% (which is probably nearer the acceptable level). How far away from that they are is anyone's guess at the moment.

[GenoMax]

Maybe ONT have learnt from PacBio's mistakes. Remember the anticlimax when you first saw PacBio data?

There is some truth to that. Unfortunately this is starting to feel like "Duke Nukem Forever".

[GW_OK]

I think they could release something and just put a giant red asterisk next to it saying this is only really, really preliminary data, really.

Any data at all would, I think, get people to cut them some slack.

[earonesty]

Maybe ONT have learnt from PacBio's mistakes. Remember the anticlimax when you first saw PacBio data?

Nanopore error rate is around 4% ~Q14, or at least it was 9 months ago.
The problem is we've all been spoilt by Illumina SBS and will see anything with < Q30 (or even Q20). They've gone on record saying they won't release anything until error is <1% (which is probably nearer the acceptable level). How far away from that they are is anyone's guess at the moment.

PacBio reads, at Q8, are still quite useful for scaffolding assemblies. Crazy to hold back data when there are lots of people who could make use of long Q14 reads right now.

[BadDNA]

Let us scaffold!

PacBio reads, at Q8, are still quite useful for scaffolding assemblies. Crazy to hold back data when there are lots of people who could make use of long Q14 reads right now.

I agree completely!

[DaleYuzuki]

If anyone's interested, I wrote up this post after talking with Oxford and several others at ASHG.

Really too bad that there wasn't any firm specifications or pricing.

Dale

[TonyBrooks]

PacBio reads, at Q8, are still quite useful for scaffolding assemblies. Crazy to hold back data when there are lots of people who could make use of long Q14 reads right now.

The problem is that scaffold runs for de novo are so niche in the grand scheme of things. The real money will be in resequencing, RNA-Seq, ChIP-Seq etc. As a core group, I'd say 99% of things we get asked to do fall into those categories. There's still a long way to go for ONT until it becomes a viable alternative.
Q14 is only really of value when you have a bunch of short read Q30s you can pin to the scaffold.

[ECO]

If anyone's interested, I wrote up this post after talking with Oxford and several others at ASHG.

Really too bad that there wasn't any firm specifications or pricing.

Dale

That ONT post mentioned Ion Torrent's party more than it did ONT. Heh.

[GW_OK]

Dale's post does have an insightful comment at the bottom. How much of the silence from ONT is because they don't have anything to show and how much is due to their current arbitration with Illumina? Maybe they're just playing it close to the vest?

[seqnextgen]

The problem is that scaffold runs for de novo are so niche in the grand scheme of things. The real money will be in resequencing, RNA-Seq, ChIP-Seq etc. As a core group, I'd say 99% of things we get asked to do fall into those categories. There's still a long way to go for ONT until it becomes a viable alternative.
Q14 is only really of value when you have a bunch of short read Q30s you can pin to the scaffold.

Maybe, but once de novo assembly is really available with a reasonable price, I believe it is the way to go. Ever wonder where those "copy number variants" are in a genome. It might be hard to believe that where they are is not important.

[pmiguel]

The other major claim is that no library construction (or even DNA purification) is required. That would immediately put these devices in their own league. That makes HiSeqs and Torrents niche instruments.

Of course there is the weaselly "sample needs 10 base single stranded leader" statement. Not sure what that means. Like you need a cos site digested with lambda terminase? Or you can just heat your sample a little to melt the terminii?

--
Phillip