New PacBio system: Sequel

SNPsaurus · 09-30-2015, 03:34 PM

http://blog.pacificbiosciences.com/2...-scalable.html

AllSeq · 09-30-2015, 06:51 PM

It isn't exactly the 'desktop' system we were waiting for, but it looks like they made some pretty big improvements in a number of areas. I'd be a little bummed if I had just bought an RSII.

luc · 09-30-2015, 08:27 PM

Seven times more data per SMRT-cell is promising.
Half the price of the RSII sequencer? "only" $350,000?

(... the RSII is much better looking than this fridge).

GenoMax · 10-01-2015, 03:59 AM

Big if is would it work consistently. That has been our biggest gripe with RSII.

Brian Bushnell · 10-01-2015, 04:51 AM

We have 3 RSII's, and have not had any particular problems with their reliability. Perhaps you got a lemon?

For our uses, this is pretty revolutionary - we could replace our 3 RSII's with one of these (they take up lot of space) and double our throughput while drastically reducing the per-base cost.

GenoMax · 10-01-2015, 05:01 AM

It does feel like we have a lemon.

It is possible that JGI has better control over samples/libraries. Do you only run internal samples? Being a core facility we get stuff all over the place and it has been impossible to get consistent P1 productivity. PacBio seems to have very narrow tolerances (compared to Illumina) on what makes a library good/productive.

luc · 10-01-2015, 12:14 PM

We see quite a lot of clearly reagent-related variation (e.g. several batches of SMRT-cells did require loading with enormous library concentrations this summer) .

Brian Bushnell · 10-01-2015, 01:54 PM

Quote:

Originally Posted by GenoMax

It is possible that JGI has better control over samples/libraries. Do you only run internal samples?

No... most of what we sequence comes from remote users. But, we do all the library construction here, and I believe we reject samples that do not have high molecular weight DNA, or insufficient mass.

SNPsaurus · 10-01-2015, 09:10 PM

For de novo work especially this is very appealing. For a small facility though I would worry that a large investment could be nullified by MinION/PromethION. That's always true of any technology, but the risks seem higher here.

nucacidhunter · 10-02-2015, 02:02 AM

PacBio is an innovative company and in recent years they have released new chemistry almost every year which has increased sequencing yield and length. I can’t say that about Oxford Nanopore and in fact they have been very slow to develop a product that works. With current pricing they are the most expensive platform for Gb of data. Their celebrity style marketing and promotion also suggests that they may not have much to offer in near future and they are just trying to stall competition. With recent trends that even companies obsolete their own systems after a year why should one wait for ONT?

ymc · 10-05-2015, 05:51 PM

This box sounds exciting. Any words on reagent cost, accuracy and read length?

AllSeq · 10-05-2015, 07:56 PM

Quote:

Originally Posted by ymc

This box sounds exciting. Any words on reagent cost, accuracy and read length?

What they've said so far:
$700/chip (including sequencing reagents) for ~7Gb of output (5-10Gb)
Read length at launch will be 8-12kb (and then increase over time).
I'm not sure if they said anything about accuracy.

luc · 10-06-2015, 05:45 PM

They said the chemistry is identical to the RSII - basically it is a bigger SMRT-cell and optimized camera and computing at the moment. Other than the read numbers the specs are likely unchanged.
For bacterial sequencing the RSII might be preferable under some circumstances, since it already generates surplus data and running a SMRT-cell on the Sequel will likely cost twice as much.

gringer · 10-06-2015, 06:41 PM

Quote:

Originally Posted by AllSeq

What they've said so far:
$700/chip (including sequencing reagents) for ~7Gb of output (5-10Gb)
Read length at launch will be 8-12kb (and then increase over time).
I'm not sure if they said anything about accuracy.

This is very close to the current cost of MinION sequencing. It wouldn't surprise me if that weren't a coincidence.

A great MinION run will currently put out about 1Gbp of sequence, but that will change substantially after fast mode kicks in (about 20x sequencing speed). I wonder how flexible PacBio are with their pricing for the chip and reagents.

AllSeq · 10-06-2015, 07:21 PM

Quote:

Originally Posted by gringer

I wonder how flexible PacBio are with their pricing for the chip and reagents.

My guess (with no evidence to back it up) is that PacBio won't directly compete with the MinION until ONT demonstrates that it can achieve the same level of de novo consensus accuracy. Until then, PacBio will just focus on the high quality sequence they generate and 'let' ONT have the portable market. If ONT can match the data quality, then PacBio could be in real trouble.

gringer · 10-06-2015, 07:31 PM

What is the current consensus accuracy for PacBio?

AllSeq · 10-06-2015, 07:38 PM

PacBio claims >99.999% (which I think is based on 50X coverage) due to their random errors. If ONT has random errors (or can get 'differently biased' errors from more than one pore type), then they should be able to have a very high consensus error rate as well. But I still haven't seen a clear claim of a random error model from ONT.

gringer · 10-06-2015, 07:43 PM

Consensus is not that good with ONT at the moment, mostly because the base calling model has consistency issues with particular pentamers (with AAAAA being a particularly good/poor example) -- we were only able to manage 99% with our Influenza genome.

I still maintain that this is a software issue, and that by throwing enough excited programmers at a proper base caller, we'll be able to re-call old reads and demonstrate that the accuracy was there all along.

Brian Bushnell · 10-06-2015, 08:20 PM

I think the most important advantage of PacBio over Nanopore is the ability to do consensus for individual reads. This is fundamental and will not change with future iterations of either platform.

There is no single "consensus accuracy" for PacBio. For all-to-all mapping, the more data, the more accurate... and for shorter inserts, the more passes, the more accurate. Considering reads of insert, which are only corrected via self-consensus, the longer the raw read length, the more accurate it will be. And the raw length is increasing rapidly with new chemistry advances.

Both Nanopore and PacBio can do all-to-all mapping for error-correction. In this situation PacBio has a huge advantage in that the errors are more randomly distributed. But Nanopore can only do self-correction with 2 passes. PacBio can do self-correction with an arbitrary number of passes, proportional to the movie length versus insert size. Therefore, things like accurate full-length 16S sequencing are already possible for PacBio, but will never be possible for Nanopore.

gringer · 10-06-2015, 08:30 PM

Isn't that just a chemistry thing though? PacBio is a sequencing-by-synthesis method after all, and I don't see any reason why Circular Consensus Synthesis couldn't be done on nanopore as well by generating the sequence prior to putting it through the sequencer.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
PacBio Library Prep workshop and PacBio SMART-Portal bootcamp - UC Davis - April 2015	DNATECH	Events / Conferences	1	04-02-2015 09:33 AM
Pacbio scaffolding	AdrianP	Pacific Biosciences	21	03-30-2015 10:09 AM
pacbio sequencing	CPCantalapiedra	Pacific Biosciences	2	11-22-2012 01:07 PM
RNAseq by PacBio	Ddhamburger	RNA Sequencing	2	06-29-2012 04:17 AM
Pacbio C2 chemistry	dongyongdong	Pacific Biosciences	3	11-15-2011 02:34 AM

09-30-2015, 03:34 PM	#1
SNPsaurus Registered Vendor Location: Eugene, OR Join Date: May 2013 Posts: 521	New PacBio system: Sequel http://blog.pacificbiosciences.com/2...-scalable.html __________________ Providing nextRAD genotyping and PacBio sequencing services. http://snpsaurus.com

09-30-2015, 06:51 PM	#2
AllSeq Registered Vendor Location: San Diego, CA Join Date: Oct 2013 Posts: 138	It isn't exactly the 'desktop' system we were waiting for, but it looks like they made some pretty big improvements in a number of areas. I'd be a little bummed if I had just bought an RSII. __________________ AllSeq - The Sequencing Marketplace info@AllSeq.com www.AllSeq.com

10-01-2015, 09:10 PM	#9
SNPsaurus Registered Vendor Location: Eugene, OR Join Date: May 2013 Posts: 521	For de novo work especially this is very appealing. For a small facility though I would worry that a large investment could be nullified by MinION/PromethION. That's always true of any technology, but the risks seem higher here. __________________ Providing nextRAD genotyping and PacBio sequencing services. http://snpsaurus.com

10-06-2015, 07:38 PM	#17
AllSeq Registered Vendor Location: San Diego, CA Join Date: Oct 2013 Posts: 138	PacBio claims >99.999% (which I think is based on 50X coverage) due to their random errors. If ONT has random errors (or can get 'differently biased' errors from more than one pore type), then they should be able to have a very high consensus error rate as well. But I still haven't seen a clear claim of a random error model from ONT. __________________ AllSeq - The Sequencing Marketplace info@AllSeq.com www.AllSeq.com

09-30-2015, 08:27 PM	#3
luc Senior Member Location: US Join Date: Dec 2010 Posts: 453	Seven times more data per SMRT-cell is promising. Half the price of the RSII sequencer? "only" $350,000? (... the RSII is much better looking than this fridge).

10-01-2015, 03:59 AM	#4
GenoMax Senior Member Location: East Coast USA Join Date: Feb 2008 Posts: 7,087	Big if is would it work consistently. That has been our biggest gripe with RSII.

10-01-2015, 04:51 AM	#5
Brian Bushnell Super Moderator Location: Walnut Creek, CA Join Date: Jan 2014 Posts: 2,707	We have 3 RSII's, and have not had any particular problems with their reliability. Perhaps you got a lemon? For our uses, this is pretty revolutionary - we could replace our 3 RSII's with one of these (they take up lot of space) and double our throughput while drastically reducing the per-base cost.

10-01-2015, 05:01 AM	#6
GenoMax Senior Member Location: East Coast USA Join Date: Feb 2008 Posts: 7,087	It does feel like we have a lemon. It is possible that JGI has better control over samples/libraries. Do you only run internal samples? Being a core facility we get stuff all over the place and it has been impossible to get consistent P1 productivity. PacBio seems to have very narrow tolerances (compared to Illumina) on what makes a library good/productive.

10-01-2015, 12:14 PM	#7
luc Senior Member Location: US Join Date: Dec 2010 Posts: 453	We see quite a lot of clearly reagent-related variation (e.g. several batches of SMRT-cells did require loading with enormous library concentrations this summer) .

10-02-2015, 02:02 AM	#10
nucacidhunter Jafar Jabbari Location: Melbourne Join Date: Jan 2013 Posts: 1,238	PacBio is an innovative company and in recent years they have released new chemistry almost every year which has increased sequencing yield and length. I can’t say that about Oxford Nanopore and in fact they have been very slow to develop a product that works. With current pricing they are the most expensive platform for Gb of data. Their celebrity style marketing and promotion also suggests that they may not have much to offer in near future and they are just trying to stall competition. With recent trends that even companies obsolete their own systems after a year why should one wait for ONT?

10-05-2015, 05:51 PM	#11
ymc Senior Member Location: Hong Kong Join Date: Mar 2010 Posts: 498	This box sounds exciting. Any words on reagent cost, accuracy and read length?

10-06-2015, 05:45 PM	#13
luc Senior Member Location: US Join Date: Dec 2010 Posts: 453	They said the chemistry is identical to the RSII - basically it is a bigger SMRT-cell and optimized camera and computing at the moment. Other than the read numbers the specs are likely unchanged. For bacterial sequencing the RSII might be preferable under some circumstances, since it already generates surplus data and running a SMRT-cell on the Sequel will likely cost twice as much.

10-06-2015, 07:31 PM	#16
gringer David Eccles (gringer) Location: Wellington, New Zealand Join Date: May 2011 Posts: 838	What is the current consensus accuracy for PacBio?

10-06-2015, 07:43 PM	#18
gringer David Eccles (gringer) Location: Wellington, New Zealand Join Date: May 2011 Posts: 838	Consensus is not that good with ONT at the moment, mostly because the base calling model has consistency issues with particular pentamers (with AAAAA being a particularly good/poor example) -- we were only able to manage 99% with our Influenza genome. I still maintain that this is a software issue, and that by throwing enough excited programmers at a proper base caller, we'll be able to re-call old reads and demonstrate that the accuracy was there all along.

10-06-2015, 08:20 PM	#19
Brian Bushnell Super Moderator Location: Walnut Creek, CA Join Date: Jan 2014 Posts: 2,707	I think the most important advantage of PacBio over Nanopore is the ability to do consensus for individual reads. This is fundamental and will not change with future iterations of either platform. There is no single "consensus accuracy" for PacBio. For all-to-all mapping, the more data, the more accurate... and for shorter inserts, the more passes, the more accurate. Considering reads of insert, which are only corrected via self-consensus, the longer the raw read length, the more accurate it will be. And the raw length is increasing rapidly with new chemistry advances. Both Nanopore and PacBio can do all-to-all mapping for error-correction. In this situation PacBio has a huge advantage in that the errors are more randomly distributed. But Nanopore can only do self-correction with 2 passes. PacBio can do self-correction with an arbitrary number of passes, proportional to the movie length versus insert size. Therefore, things like accurate full-length 16S sequencing are already possible for PacBio, but will never be possible for Nanopore.

10-06-2015, 08:30 PM	#20
gringer David Eccles (gringer) Location: Wellington, New Zealand Join Date: May 2011 Posts: 838	Isn't that just a chemistry thing though? PacBio is a sequencing-by-synthesis method after all, and I don't see any reason why Circular Consensus Synthesis couldn't be done on nanopore as well by generating the sequence prior to putting it through the sequencer.