SEQanswers

Go Back   SEQanswers > Sequencing Technologies/Companies > The Pipeline



Similar Threads
Thread Thread Starter Forum Replies Last Post
"allele balance ratio" and "quality by depth" in VCF files efoss Bioinformatics 2 10-25-2011 11:13 AM
Relatively large proportion of "LOWDATA", "FAIL" of FPKM_status running cufflink ruben6um Bioinformatics 3 10-12-2011 12:39 AM
The position file formats ".clocs" and "_pos.txt"? Ist there any difference? elgor Illumina/Solexa 0 06-27-2011 07:55 AM
"Systems biology and administration" & "Genome generation: no engineering allowed" seb567 Bioinformatics 0 05-25-2010 12:19 PM
SEQanswers second "publication": "How to map billions of short reads onto genomes" ECO Literature Watch 0 06-29-2009 11:49 PM

Reply
 
Thread Tools
Old 02-17-2012, 11:58 AM   #21
SeqAA
Guest
 

Posts: n/a
Default

definitely more useful than i was anticipating. I'm still unsure about a few parts. At least they are upfront about the raw error rate which wasn't too bad. The low read count is still concerning with such an error rate, but it makes a great companion for hybrid datasets.
  Reply With Quote
Old 02-17-2012, 07:18 PM   #22
BBoy
Member
 
Location: Pacific Northwest

Join Date: Oct 2010
Posts: 52
Default

So let me run a score card on my own prognosticaltion

Quote:
Originally Posted by BBoy View Post
My predictions: long read lengths, on par or beyond what PacBio can achieve;
Check, though the claimed read length is massively higher than Pac Bio

Quote:
Originally Posted by BBoy View Post
raw accuracy better than PacBio, but probably still in the low-mid 90s;
Check check

Quote:
Originally Posted by BBoy View Post
poor data output per rack, with high consumable costs due to relatively low multiplexing per chip
Yes and no. The raw cost per base is apparently low, but the multiplexing is indeed presently very low - hundreds to a few thousand wells. This is the main issue I see for them - with solid state technology being where it is they need to rely on biological pores to get the resolution and therefore the outlook for massive parallelism is somewhat murky.

Overall impressive technical results, I need to disgest them some over the weekend. Seems more like a threat to the PGM than PacBio at first glance - simple sequencing at a low capital cost, with what seems to be much lower reagent consumption and simpler sample prep.
BBoy is offline   Reply With Quote
Old 02-17-2012, 07:46 PM   #23
ECO
--Site Admin--
 
Location: SF Bay Area, CA, USA

Join Date: Oct 2007
Posts: 1,352
Default

Quote:
Originally Posted by BBoy View Post
Define "easier". Easier to achieve higher raw accuracy... yes. Easier to achieve reasonable multiplexing... not only no, but hell no. At least not for as long as they are using biological nanopores.
Can't believe I missed this...tried to resist replying.

Multiplexing for PacBio requires real time multi-axis nanometer-scale alignment of multiple excitation lasers and emission collection equipment across a macroscopic (~4mm circle for 2x75,000) array of ZMWs that also has to be machined/synthesized optically flat. This is an optics problem that seems to shame the Hubble telescope.

Multiplexing of a biological nanopore array requires coating an array of sensors with biological-esque membrane and depositing one (and only one) pore atop each sensor (this is ultimately a statistics problem). I have seen data from nanopore companies that suggests this is (relatively) trivial. Then await the stream of voltage data and hope your pore can discriminate bases or groups of bases. Oh and you need a Transmeta processor and RAMBUS RAM too...!

In my mind (and perhaps slanted/naive perspective), there is no contest in the difficulty of methods. But difficulty is not much more than a badge of honor in this business, then again PacBio has a functional instrument and ONT has years of promises and two phases of vaporware launches.

Biological nanopores are a probably not a passing fad...likely the world will need many years of fab technique development before a solid-state nanopore can replicate the precision of mspA, aHL, or whatever secret pore ONT is using (likely a mutant of mspA). As far as I know there is no angstrom/atom level 3D silicon fabrication technique out there that works in huge parallel.

Again...I (perhaps obviously) have limited knowledge of anything other than molecular biology. I could be very far off with the above...but it sure is fun to speculate about!

Bottom line in my book...I have _never_ been blown away by a sequencing announcement. Everything ends up being hype followed by incremental increases in [RL/Tput/Accuracy/sampleprepsimplicity].
ECO is offline   Reply With Quote
Old 02-17-2012, 08:29 PM   #24
BBoy
Member
 
Location: Pacific Northwest

Join Date: Oct 2010
Posts: 52
Default

Quote:
Originally Posted by ECO View Post
Multiplexing for PacBio requires real time multi-axis nanometer-scale alignment of multiple excitation lasers and emission collection equipment across a macroscopic (~4mm circle for 2x75,000) array of ZMWs that also has to be machined/synthesized optically flat. This is an optics problem that seems to shame the Hubble telescope.
Yes, this was an unfortunate design decision in their instrument design that is relatively easily correctable. Fabricating an integrated waveguide distribution system that delivers light to the ZMWs and is locally aligned with sub-um precision is a trivial task by today's technology standards - every microprocessor today does that for billions of features with a precision of a few nanometers. Slapping a dedicated disposable CMOS active pixel as the local sensing element is also a well understood task. None of these are conceptual limitations. Whether Pac Bio gets to address them before they run out of cash is an open question.

Quote:
Originally Posted by ECO View Post
Multiplexing of a biological nanopore array requires coating an array of sensors with biological-esque membrane and depositing one (and only one) pore atop each sensor (this is ultimately a statistics problem). I have seen data from nanopore companies that suggests this is (relatively) trivial. Then await the stream of voltage data and hope your pore can discriminate bases or groups of bases.
Agreed. But how do you pack millions of these on a sub-1um pitch in an ordered array that is aligned to the sensor underneath? Please don't tell me "directed self assembly"

Quote:
Originally Posted by ECO View Post
In my mind (and perhaps slanted/naive perspective), there is no contest in the difficulty of methods.
And neither is there one in my mind. The technology that solves PacBio's problem was developed decades ago and is in massive use today (every cell phone camera basically implements it). The technology required to create an ordered array of millions of biological nanopores exists only in Nature papers :-)

Quote:
Originally Posted by ECO View Post
Biological nanopores are a probably not a passing fad...likely the world will need many years of fab technique development before a solid-state nanopore can replicate the precision of mspA, aHL, or whatever secret pore ONT is using (likely a mutant of mspA). As far as I know there is no angstrom/atom level 3D silicon fabrication technique out there that works in huge parallel.
Hey, people are publishing papers on graphene for nanopore sensing. Talk about putting lipstick on a pig. Yes, synthetic nanopores are with the required dimensions and tolerance are at least a decade away. Biological nanopores are here to stay.


Quote:
Originally Posted by ECO View Post
Again...I (perhaps obviously) have limited knowledge of anything other than molecular biology. I could be very far off with the above...but it sure is fun to speculate about!
And I have limited knowledge of anything other than semiconductors. So we have a deaf communicating with a mute and a blind (now THAT was a funny movie back in its day :-)

Last edited by BBoy; 02-18-2012 at 01:14 AM.
BBoy is offline   Reply With Quote
Old 02-18-2012, 04:06 AM   #25
gringer
David Eccles (gringer)
 
Location: Wellington, New Zealand

Join Date: May 2011
Posts: 778
Default

I noticed the video suggested the nanopores can be adapted for proteins. Does this mean there'll be a single-molecule protein sequencing pipeline available in the near future?

http://www.nanoporetech.com/technolo...gridion-system
gringer is offline   Reply With Quote
Old 02-18-2012, 05:51 AM   #26
nickloman
Senior Member
 
Location: Birmingham, UK

Join Date: Jul 2009
Posts: 356
Default

My understanding from what they told me is that there isn't a 1:1 relationship between circuits and pores. In fact there are more many circuits than pores, and a pore can address between 0 and 4 circuits. Presumably they have a method for sorting out the cross-talk (sorry not to be more specific).
nickloman is offline   Reply With Quote
Old 02-18-2012, 06:16 AM   #27
Geneus
Member
 
Location: New Jersey

Join Date: Dec 2010
Posts: 60
Default

Quote:
Originally Posted by BBoy View Post
Yes, this was an unfortunate design decision in their instrument design that is relatively easily correctable. Fabricating an integrated waveguide distribution system that delivers light to the ZMWs and is locally aligned with sub-um precision is a trivial task by today's technology standards - every microprocessor today does that for billions of features with a precision of a few nanometers. Slapping a dedicated disposable CMOS active pixel as the local sensing element is also a well understood task. None of these are conceptual limitations. Whether Pac Bio gets to address them before they run out of cash is an open question.



Agreed. But how do you pack millions of these on a sub-1um pitch in an ordered array that is aligned to the sensor underneath? Please don't tell me "directed self assembly"


And neither is there one in my mind. The technology that solves PacBio's problem was developed decades ago and is in massive use today (every cell phone camera basically implements it). The technology required to create an ordered array of millions of biological nanopores exists only in Nature papers :-)


Hey, people are publishing papers on graphene for nanopore sensing. Talk about putting lipstick on a pig. Yes, synthetic nanopores are with the required dimensions and tolerance are at least a decade away. Biological nanopores are here to stay.



And I have limited knowledge of anything other than semiconductors. So we have a deaf communicating with a mute and a blind (now THAT was a funny movie back in its day :-)

Ouch...this stuff is waaaaaay over my head. I'll need to ask my brother to read and decipher (he's Chief Systems Technologist at a large chip foundry). Who knows, maybe I can peak his interest.
Geneus is offline   Reply With Quote
Old 02-18-2012, 06:36 AM   #28
ECO
--Site Admin--
 
Location: SF Bay Area, CA, USA

Join Date: Oct 2007
Posts: 1,352
Default

Quote:
Originally Posted by BBoy View Post
Yes, this was an unfortunate design decision in their instrument design that is relatively easily correctable. [snip]

And I have limited knowledge of anything other than semiconductors. So we have a deaf communicating with a mute and a blind (now THAT was a funny movie back in its day :-)
Awesome post. I can't say anymore about waveguide distribution...this would be a fun conversation over a beer.
ECO is offline   Reply With Quote
Old 02-18-2012, 07:47 AM   #29
pmiguel
Senior Member
 
Location: Purdue University, West Lafayette, Indiana

Join Date: Aug 2008
Posts: 2,195
Default

The first thing I would want to do would be to gather 3x coverage of 50kb+ reads from a moderate to large eukaryotic genome. (1 gigabase +). Preferably something already considered "sequenced". For example, maize, soybean or some mammal. I am hazy on the costs of doing this on a GridIon, but for soybean on a few MinIons it looks like it would cost about $3000. If there do not seem to be any major issues I would think most of the thorny issues involved in de novo genome sequencing disappear.

1 Gbp genome -- go 50x with Illumina PE reads (maybe 3 HiSeq lanes). Don't bother with the ME reads, making those libraries takes real effort. Then assemble on a scaffold of (Min|Grid)Ion reads (50 - 100 kb, right?). Instant chromosome sequences? Maybe you still have some issues with centromeric heterochromatin. But with 100 kb reads, maybe not. That is 1 month project you can do for $20K, right?

So I am reading right that they are only releasing phage lambda data though? Or are they even releasing that? My experience is that even E. coli Dam methylation is incomplete in lambda clones. Anyone seeing an increase in errors at Dam (GATC) sites?

Another potential issue: it has never been clear to me what level of damage exists in a typical DNA prep. Seems like oxidation/adduct/apurinic/pyrimidine-dimer sites might be quite common and we would not notice it using normal (colligative) molecular techniques.

Even if this were the case, passaging the DNA of our species to be de novo sequenced through a non-modifying host would probably only double the cost of a new genome sequence. (Eg, make a cosmid library in a modification minus host strain.) Or, if you just end up with an "N" at the position of a modified or damaged base -- probably could live with that.

--
Phillip
pmiguel is offline   Reply With Quote
Old 02-18-2012, 08:28 AM   #30
ECO
--Site Admin--
 
Location: SF Bay Area, CA, USA

Join Date: Oct 2007
Posts: 1,352
Default

Quote:
Originally Posted by pmiguel View Post
Another potential issue: it has never been clear to me what level of damage exists in a typical DNA prep. Seems like oxidation/adduct/apurinic/pyrimidine-dimer sites might be quite common and we would not notice it using normal (colligative) molecular techniques.
This was an interesting problem for sample prep at PacBio, as we used several exonucleases to remove non-covalently closed SMRTbells (which also cut at abasic sites).

We encountered significant internal damage (exo-labile abasic sites) with inserts over a certain size from "high quality" DNA preps, and initially couldn't make any products from larger PCR products. I interpreted this as there were >1 abasic sites in every insert molecule larger than a certain size.

Just one of the many fun challenges of single molecule sample prep. Hopefully Oxford can already cope with real biological abasic sites (not stable THF synthetic ones) and or the adducts you mention. Basically adds additional moieties they need to detect.
ECO is offline   Reply With Quote
Old 02-19-2012, 08:07 AM   #31
pmiguel
Senior Member
 
Location: Purdue University, West Lafayette, Indiana

Join Date: Aug 2008
Posts: 2,195
Default

Quote:
Originally Posted by ECO View Post
Just one of the many fun challenges of single molecule sample prep. Hopefully Oxford can already cope with real biological abasic sites (not stable THF synthetic ones) and or the adducts you mention. Basically adds additional moieties they need to detect.
Seems like studies looking for oxidation products (8-oxo-G) always find it. Who knows how many methylation/hydroxymethylation/etc. variants are present in natural DNA samples.

Does the commercial release schedule Oxford plans seem realistic to you? I can't find any data sets they have released. So commercial release of MinIons and GridIons later this year seems fantastically aggressive to me.

In Sequence is quoting Oxford Nanopore's price per gigabase for the GridIon at around $40. $40,000/ terabase is Illumina HiSeq2500 territory 6 months from now. At that price point Illumina gives you 1-2 orders of magnitude lower error rates and that's it. Every other possible advantage would be with the GridIon.

Of course we already know the limitations and strengths of a HiSeq, whereas we have only Oxford Nanopore's marketing hype to make an assessment of their product.

Thing is, I always assumed the final state of sequencing would be a 4th gen machine that does no chemistry on the templates it sequences. That is, it would be a DNA reader. But that eventuality was over the "horizon". Might have taken us another decade to get there.

--
Phillip
pmiguel is offline   Reply With Quote
Old 02-19-2012, 09:30 AM   #32
gringer
David Eccles (gringer)
 
Location: Wellington, New Zealand

Join Date: May 2011
Posts: 778
Default

Quote:
Originally Posted by pmiguel View Post
Thing is, I always assumed the final state of sequencing would be a 4th gen machine that does no chemistry on the templates it sequences. That is, it would be a DNA reader. But that eventuality was over the "horizon". Might have taken us another decade to get there.
There are still a few more opportunities for additional generations of sequencing technology.

Consider that the sequencing still needs a lysis stage. What about in-vitro / in-vivo sequencing, or completely non-invasive sequencing?
gringer is offline   Reply With Quote
Old 02-19-2012, 09:45 AM   #33
pmiguel
Senior Member
 
Location: Purdue University, West Lafayette, Indiana

Join Date: Aug 2008
Posts: 2,195
Default

Quote:
Originally Posted by gringer View Post
There are still a few more opportunities for additional generations of sequencing technology.

Consider that the sequencing still needs a lysis stage. What about in-vitro / in-vivo sequencing, or completely non-invasive sequencing?
And plenty of work to do in the single molecule sampling in lysed samples arena. The devil is in the details.

I would say a completely non-biological sensor (solid state) would be worthy of the appellation "5th generation".

--
Phillip
pmiguel is offline   Reply With Quote
Old 02-19-2012, 10:58 AM   #34
BBoy
Member
 
Location: Pacific Northwest

Join Date: Oct 2010
Posts: 52
Default

Seems that Rothberg shares my concerns about scaleup
http://www.forbes.com/sites/matthewh...encer-a-rival/

True, he is not exactly unbiased, so this should be taken with a rock of salt.
BBoy is offline   Reply With Quote
Old 02-19-2012, 11:59 AM   #35
Pongo_T
Junior Member
 
Location: Germany

Join Date: Feb 2012
Posts: 4
Default

For the more technically minded: did Clive Brown show any primary data in the sense of simultaneously acquired nanopore current traces (there is none on the webpage, is there?) It is one thing to show that (1) strand sequencing basically works on short genomes and that (2) you have a mulitplexed nanopore chip which might be good for longer ones, but quite another to show that strand sequencing has actually successfully been done using that chip. There are lots of things where they must have made tremendous progress, not least the on-chip multiplexed 512-channel (for the Minion) voltage-clamp amplifier (CMOS-based, I understand) which has to have pretty low noise, a.s.f.

Last edited by Pongo_T; 02-19-2012 at 12:02 PM.
Pongo_T is offline   Reply With Quote
Old 02-19-2012, 12:48 PM   #36
pmiguel
Senior Member
 
Location: Purdue University, West Lafayette, Indiana

Join Date: Aug 2008
Posts: 2,195
Default

Quote:
Originally Posted by BBoy View Post
Seems that Rothberg shares my concerns about scaleup
http://www.forbes.com/sites/matthewh...encer-a-rival/

True, he is not exactly unbiased, so this should be taken with a rock of salt.
Strange, does Rothberg think the genome size of phage lambda is 5000 bases? 5x10^3 to 6x10^9 that would be six orders of magnitude. Lambda is about 48,500 bp.

Okay, I am being pedantic. It does not change his main point. Whether 5 orders of magnitude or 6 that does seem like a lot of ground to cover in less than a year. Why not sequence a bacterial genome? Or yeast? Would be really nice to see the sequence reads from a data set.

Guess we will be seeing...

--
Phillip
pmiguel is offline   Reply With Quote
Old 02-19-2012, 06:29 PM   #37
Wallysb01
Senior Member
 
Location: San Francisco, CA

Join Date: Feb 2011
Posts: 286
Default

How many times have we seen this?

Lets see some real data from a Eukaryote before we pass any kind of judgement. Heck any actual data would be a nice first step.

Its all marketing and spin until then...
Wallysb01 is offline   Reply With Quote
Old 02-19-2012, 06:41 PM   #38
aeonsim
Member
 
Location: Belgium

Join Date: Jun 2011
Posts: 45
Default

Quote:
Originally Posted by Wallysb01 View Post
How many times have we seen this?

Lets see some real data from a Eukaryote before we pass any kind of judgement. Heck any actual data would be a nice first step.

Its all marketing and spin until then...
Yes as promising as it looks we certainly need some real raw data, a lambda dataset and a Eukaryote would do much to to inspire confidence!

Though as soon as they make MinION preorders available we'll order half a dozen for testing just to get a feel for the data with our organism, and current methods. The minions really are an interesting looking product and have awesome potential as marketing tool that makes a nice profit.

If data from the minions looks good then the primary instrument becomes very tempting assuming a reasonable cost.
aeonsim is offline   Reply With Quote
Old 02-19-2012, 06:50 PM   #39
Wallysb01
Senior Member
 
Location: San Francisco, CA

Join Date: Feb 2011
Posts: 286
Default

Quote:
Originally Posted by aeonsim View Post
Yes as promising as it looks we certainly need some real raw data, a lambda dataset and a Eukaryote would do much to to inspire confidence!

Though as soon as they make MinION preorders available we'll order half a dozen for testing just to get a feel for the data with our organism, and current methods. The minions really are an interesting looking product and have awesome potential as marketing tool that makes a nice profit.

If data from the minions looks good then the primary instrument becomes very tempting assuming a reasonable cost.
It certain looks cool. Sequence a genome in the palm of your hand! Who isn't willing to at least try that? And if it actually works maybe those little vacuum cleaners from GATACA that sequence your DNA in a few seconds will be possible! But really, who's going to get paper print out of such a thing....
Wallysb01 is offline   Reply With Quote
Old 02-20-2012, 02:43 PM   #40
ECO
--Site Admin--
 
Location: SF Bay Area, CA, USA

Join Date: Oct 2007
Posts: 1,352
Default

Quote:
Originally Posted by aeonsim View Post
Though as soon as they make MinION preorders available
If they are available in 2012, I would love to arrange for a SEQanswers community beta test of 10 of them...even if I have to hit my savings account to pay for them.

Unfortunately beta testers these days are under lock-tight NDA's so they can't share their real experiences. Chance of the above happening? You decide.
ECO is offline   Reply With Quote
Reply

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off




All times are GMT -8. The time now is 08:30 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2017, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO