New Tech at AGBT

[aeonsim]

I thought it might be useful to collate the bits and pieces about new platforms discussed during AGBT (Those that don't already have a thread), for those of us unlucky enough not to attended. Grabbing the info from various tweets listed at: http://pathogenomics.bham.ac.uk/blog...-day-2-tweets/

If any one at AGBT could supply additional information from the talks, that'd be appreciated or any links to blogs covering the tech.

GnuBIO

Read Length: 1000bp
Accuray: Phred 70+
Holopolymers: Phred 70+ to 9bp
Sample Prep: < 1min

System is rackable, ~30lbs, Dry Instrument fludics in cartridge. Beta Mid-2012, commercial end of 2012? Real time Variant calling, each base queried 6 times.

Source:
http://t.co/JPKbR7L2
http://t.co/5KSzCcVS

Technique/Chemistry: ?
Website: http://gnubio.com/

LaserGen

Raw base accuracy 99.8%
100bp reads
~$99K for instrument?
$1K per run?

Technique/Chemistry:?
Website: http://lasergen.com/

So what do people think, do they sound interesting or is it all irrelevant and they'll be buried under ONTs upcoming announcement?

[Geneus]

Originally Posted by aeonsim

So what do people think, do they sound interesting or is it all irrelevant and they'll be buried under ONTs upcoming announcement?

Buried.


http://finance.yahoo.com/news/Oxford...13504.html?x=0

[GW_OK]

I agree with "buried".
All equipment purchases are going to be re-thought in light of ONT...

[larissa]

Even with 4% error rates? Will they deliver in improving that? That's not acceptable for clinical use. It may be very useful for a lot of other stuff.

[Geneus]

Originally Posted by GW_OK

I agree with "buried".
All equipment purchases are going to be re-thought in light of ONT...

ILMN and LIFE are selling off today...I wonder why?

[larissa]

Originally Posted by Geneus

ILMN and LIFE are selling off today...I wonder why?

So, doesn't Illumina have stakes in ONT?

[Geneus]

Originally Posted by larissa

So, doesn't Illumina have stakes in ONT?

Yes...an investment. But apparently they (ILMN) will not get access to these two new platforms from Oxford. Too bad for them.

[nxgsqq]

No way Illumina will take this lying down. I'll bet we'll hear major announcements from them. They have a track record of coming from behind and taking over, let's see if they still have the hunger to do it. I only see this as the field opening up to truly next generation technologies.

[larissa]

They better do something with Roche coming after them.

[Geneus]

Originally Posted by larissa

They better do something with Roche coming after them.

Maybe Roche will drop their pursuit of ILMN and go after ONT...now that would be BIG news.

[aeonsim]

Yes it does look like they maybe buried now. Unless they can find something that carves them out a niche in the market that nanopores can't currently target. I see the odd comment suggesting that the laser Gen chemistry might be compatible with illuminas tech if it's an improvement to there chemical it could end up being purchased.

I wonder if illumina will announce a 1.2tb a run upgrade for the Hiseq and compete on data volume and higher accuracy.

[gringer]

Originally Posted by larissa

Even with 4% error rates? Will they deliver in improving that? That's not acceptable for clinical use. It may be very useful for a lot of other stuff.

If this is a raw per-molecule error (as I assume with a "single molecule" technology), then I consider it acceptable. Illumina and SOLiD technologies use hundreds of identical sequencing reactions per cluster to generate a consensus read, and it's unlikely they would have the same accuracy with single molecules per cluster.

It has been mentioned that the errors are almost entirely deletion errors. It would be possible to sequence more molecules and do a fairly simple alignment process on the sequences to identify deletions and get a consensus.

Assuming the error is random enough (bearing in mind the press releases suggest there is some bias), multiple sequencing experiments should increase the chance of a reliable read fairly quickly. With two strands sequenced (e.g. forward and reverse) and completely random error, you double the Q value (i.e. from ~Q14 / p = 0.04 to ~Q28 / p = 0.0016). The bias would increase the chance of two strands having an error at the same location, which would probably mean there'll be some tricky regions that still can't be sequenced accurately, regardless of how many times it is done in parallel.

Bear in mind that tricky regions are already accepted as a fact of life with previous sequencing technology. In particular, very long highly repetitive regions, or regions that are hypervariable within individuals are a problem with current technology.

[Geneus]

Originally Posted by gringer

If this is a raw per-molecule error (as I assume with a "single molecule" technology), then I consider it acceptable. Illumina and SOLiD technologies use hundreds of identical sequencing reactions per cluster to generate a consensus read, and it's unlikely they would have the same accuracy with single molecules per cluster.

It has been mentioned that the errors are almost entirely deletion errors. It would be possible to sequence more molecules and do a fairly simple alignment process on the sequences to identify deletions and get a consensus.

Assuming the error is random enough (bearing in mind the press releases suggest there is some bias), multiple sequencing experiments should increase the chance of a reliable read fairly quickly. With two strands sequenced (e.g. forward and reverse) and completely random error, you double the Q value (i.e. from ~Q14 / p = 0.04 to ~Q28 / p = 0.0016). The bias would increase the chance of two strands having an error at the same location, which would probably mean there'll be some tricky regions that still can't be sequenced accurately, regardless of how many times it is done in parallel.

Bear in mind that tricky regions are already accepted as a fact of life with previous sequencing technology. In particular, very long highly repetitive regions, or regions that are hypervariable within individuals are a problem with current technology.

Excellent synopsis...thank you.

[nxgsqq]

I too agree the error is acceptable, IF, it is indeed the typical error they see and not a 'best case' presented for effect and advertisement. I am suspicious of resolving 64 levels (3 base read) electronically considering how small the differences will be. I don't completely buy the algorithmic deconvolution either, especially if they are still using a polymerase. If it is a non-stochasitic transport, a Viterbi/HMM algorithm might give 94% accuracy.

The bigger unknown is the true, customer usability of their pores. I would assume they are Poisson loading the pores before they ship to users. How many pores are still active after an hour of use? I know the bilayers can be made stable and inert, I can accept the error profile can be made length independent (especially if they tether the motor to the pore, else Brownian motion of long DNA can act to pull the complex off, even against the electric field), but how many pores are sequening at a given time and how does that number drop off over time?

There is plenty of smoke and mirrors happening, but to give credit where it is due, kudos to them for enabling, even a semblance of, a product around nanopores, and for a dignified and sombre presence.

[nickloman]

My understanding is that it's not a polymerase, they are calling it a "special progressive enzyme".

[clivey]

Originally Posted by gringer

If this is a raw per-molecule error (as I assume with a "single molecule" technology), then I consider it acceptable. Illumina and SOLiD technologies use hundreds of identical sequencing reactions per cluster to generate a consensus read, and it's unlikely they would have the same accuracy with single molecules per cluster..

Thank you for spotting that, it's nice to meet someone with insight.

C.

[Heisman]

Originally Posted by gringer

The bias would increase the chance of two strands having an error at the same location, which would probably mean there'll be some tricky regions that still can't be sequenced accurately, regardless of how many times it is done in parallel.

You sure about this? With the complement strand the sequencer would see not only a completely different series of bases, but bases from the "other" direction, I think. if anything I think the error profile would be remarkably different for complementary strands.

[Chipper]

Originally Posted by clivey

Thank you for spotting that, it's nice to meet someone with insight.

C.

Was the 4% error rate for the consensus sequence from hairpin reads or for single strands?

[gringer]

Originally Posted by nxgsqq

The bigger unknown is the true, customer usability of their pores. I would assume they are Poisson loading the pores before they ship to users.

The information on the website suggests that the pores (in solution) are loaded onto the chip at the start of a run:

http://www.nanoporetech.com/technolo...ge/sensor-chip

The sensor chips are stored dry, but on starting the experiment, the flow cell is automatically exposed to the relevant electrophysiological and other fluids required to create nanopores in bilayers in experimental conditions.

I presume the system will be set up so that the reported shelf life of the cartridges will give a reasonable proportion of active nanopores up to the use by date.

They intend to use solid-state pores in the future, which I guess would make the cartridges last a whole lot longer.

http://www.nanoporetech.com/technolo...tate-nanopores

How many pores are still active after an hour of use? I know the bilayers can be made stable and inert, I can accept the error profile can be made length independent (especially if they tether the motor to the pore, else Brownian motion of long DNA can act to pull the complex off, even against the electric field), but how many pores are sequencing at a given time and how does that number drop off over time?

The occupancy/speed is customisable. Faster movement has a quality reduction and greater chance of missed bases. From reading one of the press releases (I forget which one), it sounded like the USB MinION would run for no more than 10 hours, but the GridION cartridges could keep going for a few days. Samples can be ejected during a run and shifted to other cartridges, cartridges can be replaced, runs can be started at any time, and stopped due to a number of different desired factors, so any drop-off in occupancy probably won't have too much impact on result output (as long as there's still money to burn).

[gringer]

Originally Posted by Heisman

You sure about this? With the complement strand the sequencer would see not only a completely different series of bases, but bases from the "other" direction, I think. if anything I think the error profile would be remarkably different for complementary strands.

I can only guess at the nature of the error that they are talking about, but the "tricky regions" may not necessarily be an obvious class of sequences. Maybe there'll be an issue with only A/T homopolymers advancing a bit slower (or quicker) than expected. Perhaps there would be particular palindromic sequences (appearing the same in reverse complement) that have a high likelihood of read errors even when sequencing both strands.

Even if it would be possible to reduce error by doing more sequencing, they will still need to try their hardest to increase the single-strand accuracy (and reduce bias) given that it's reported as a single-molecule sequencing technology.