[jdelton]

The 250gb is what is being promised with the P3 chip which is scheduled for release I believe the end of the year. I haven't used the V2 chemistry yet so I'm still at 170, but once we've used up the current kits we'll move forward. And no not a paired end read, just a single read. The cost of a proton is more than the PGM, but not much more than the MiSeq (though the peripherals move it up to almost double the MiSeq), and the run cost is comparable too. The size of the machine is benchtop as well (at least it is in our lab), it's only slightly wider than a PGM, but about 1.5x as deep. I'm probably going to eat my words but right now I have a lot of hope for the Life machines.

[jdelton]

It very well may be that the Proton was an answer to HiSeq, in our lab we see it as an alternative to MiSeq, primarily because of the data limit of PGM, but also because the upgrades have not been released (chip and chemistry) yet to allow proton to compete very effectively with HiSeq.

[Buzz0r]

Good morning,
this is actually entertaining to read with my morning cafe latte. You know you sound like Rothberg 1 year ago (actually this could be copy paste from one of their flyers probably). Clearly their marketing has done you well, but I can't understand the naive comments you posted here, but let me be more specific as I am not trying to criticize your experience (which obviously is going very well with Ion Torrent for you personally!), just giving you some food for thoughts:

- PGM and Proton not mature?: Proton is not mature at all as it's output is way below promised specs and what we see in the APAC countries is very poor performance throughout most Proton owners (including DNAlink who are like a Proton Poster Boy, still cant get any proper runs done!). PGM on the contrary is actually pretty much at it's maximum, scaling up doesn't look to be continuing as before and the longer read length cause so much problems on the PGM already that porting it to Proton is looking even more difficult.

- Miseq maxed out?: Miseq is far away from it's maximum, 2x300Bp commercial kits coming out soon, 2x400 runs done internally and if you like to talk to Dr Schuster from UPenn, he can show you runs he did in his lab overlapping paired reads to around 500Bp with current 2x250 kits, getting highest quality data. Read numbers are also going up on the Miseq as the flow cells will be enhanced by a different clustering method so I am not sure what you are talking about limited.

- 400Bp easy on PGM?: Well you must be one lucky person to get this done easily as we do not see the results you are talking about at all. Maximum read length maybe, but average is far away from that. Not sure if you are interested in the maximum of some reads 500Bp if your average is at 320Bp with a quality that sucks compared to what the Miseq is delivering (Q30 plus compared to lousy Q17!).

- Getting over specs: Yes for 314/316! 318 is tough to even get to the specs, not sure on the new chips but getting to 1Gb on a 318 is already quite a challenge!

- Proton specs galore: 200Bp on Proton? Up to 4 weeks ago IT tech guys still said not to run the 200Bp kits as they are not working well, so we see 100Bp runs at average ~83Bp, 50% raw data at a quality of Q20 or higher. New kits are being tested at the moment for 200Bp (finally as it took them almost a year to get this problem fixed which was promised early on). I do not see any indication that this will jump to 400 Bp any time soon with all the problems occurring and the time they took just to get from 100 to 200! Maybe want to get to 300 first before announcing 600Bp!

- New Avalanche and Chef: I assume people will have a company celebration once this really takes place! Release dates have been pushed over and over, IT guys don't even know if it really will come out this year or mid next year or the year after? Are we in 2016 yet where all this might be completely redundant because we are doing nanopores? Not hitting a single milestone in time for the past 18 months just does not give me faith in new stuff anymore. Also it is not solving all technology related problems like homoplymer issues, accuracy and so on...if you can't fix the indel problem you might as well just move on and develop some new technology as this just limits your application tremendously and excludes you from the clinical market completely.

-250Gb per chip: not sure how you put this into any context here as the P3 has been completely pushed back and most likely is not ready until 2015 (at least this is the date they present now since "tbd" looks too crappy!) They also lowered the specs for the P3 btw to 64 Gb!

- Miseq will be surpassed quickly: do you think Illumina is not developing the technology at all? It looks to me that with any release IT is limping behind Illumina might already have something ready to go and counter. It is just ridiculous to think the MiSeq will no longer be upgraded!

- P3 end of the year?: you mixed this one up, the P2 is supposed to come out by the end of the year which basically means 1q2014, P3 is far away!

- Proton competing with Hiseq: not sure how you believe what Rothberg is praising here but even if the Proton can get their specs up in the next 3 years to a 100-200Gb output of usable data (60% above Q20 is what you can expect as everything else is for the bin bioinformatically!), Illumina will also push development for 3 years which I hedge way more promise for than IT's development.

I am glad you have good results with your PGM, not sure if you are willing to post the same for your Proton. So far no one posted a dataset of their Proton only so that speaks loud enough for me. I am alos not saying do not buy PGM or Proton but i want to make sure that expectations are appropriate and that information about technology is fact and not future. Clearly i am happy with MiSeq and HiSeq for many reasons i do not want to get into here as this is not a marketing room but a forum for experiences made by scientists and not promised by companies.

Have a good day!

Cheers T.

Cheers T.

[thomasblomquist]

Heard from a life tech rep at the last AACR meeting in DC that Avalanche was off the table. Not sure what that was supposed to mean, but it is what it is.

[Buzz0r]

...question is was it "on a table" at any time? ... tired of promises!

[thomasblomquist]

Hi,

Tom Blomquist here, University of Toledo Medical Center, OH, USA.

Just a "couple" of observations from my own experience with both the Ion Torrent as well as Illumina Miseq. As a note, I do solely amplicon type sequencing with our own in-house protocol for library prep. We then send our prepped libraries to other core-labs because it's 1) more cost-effective for us to do so, and 2) by the time we were to purchase an instrument we would be eating dirt kicked up by the next version/iteration of NGS instrumentation that comes along. Our protocol enables cross-platform direct comparisons of sequence read quality. We accomplish this by using a fusion primer PCR protocol which adds different patient specific barcode sequences (again in-house custom design), and then tack the Illumina specific or Ion Torrent specific adapters on the end. I will not give out our designs... yet. The manuscript is in review. I will after accepted, and can post appropriate links to our designs then.

My observations:

1) I like the Ion Torrent for relatively cheaper and quicker turn arounds compared to Miseq. With my amplicon libraries (100-300 targets) I can usually prep a given patient nucleic acid sample in about 6 hours with full QC and being readied for placement into either the flow chip for Miseq, or Ion Torrent emulsion PCR step for bead prep. After that, I can get back my sequencing data from Ion Torrent service later that night, or the next day from a Miseq service. Longer times apply for shipment if needed of the libraries, of course. The cost for a Ion Torrent 318 chip service in my neck of the woods when outsourced is about $1,000 1x200bp, and for Miseq, we are paying $2,500 for 2x250bp.

2) Our Miseq provider gives back ~13 million sequencing reads, of which our own QC bioinformatic filters toss out about 10% (11 million reads). Our Ion torrent provider gives back ~6.5 million sequencing reads, of which we toss out about 25% of our reads (4-5 million reads; half of miseq). There are a number of reasons for these yields: Ion torrent necessitates that you have very few to no heterodimers between your target templates during the emulsion PCR. Any sequencing beads prepared during this step with heter-templates results in a "polyclonal" bead. This bead is then filtered by the Ion Torrent Server software suite because it cannot interpret the H+ output during each flow of nucleotides. Heterodimer templates occur quite often during library prep, and occur at room temperature with the natural breathing of your prepared DNA templates. Or they can occur to library prep purification, especially if heat or any sort of chemical denaturants are involved. *** Because Illumina uses a single strand of DNA template for each colony cluster to be formed, it does not have this heterodimer issue, and thus, the majority of sequencing reads it calls as good, are indeed quite good. *** Also note, that my Ion Torrent provider for sequencing services takes pain-staking steps to optimize 318 chip loading. On average, they can get about 9.5 - 10 million beads that have been populated with template onto the 318 chip. However, the polyclonal bead issue arises and knocks it down to 6.5 million non-polyclonal beads. Thus, the cost per sequencing read is about the same between Ion Torrent 318 and Miseq. With Ion Torrent beating out in turn around by about 24 hours. For us, we are leaning toward Ion Torrent for acute panels in our patient populations, and Miseq for other less acute panels because the library prep is less stringent for heterodimerizing of templates.

*** I should note that it seems the more complex our amplicon panels get, the more heterodimerization we observe and thus diminished sequencing reads.

3) Ion torrent is not so good if you are looking for discrimination of homopolymer variants. After about 3-4 nucleotide repeats, Ion Torrent has difficulty in estimating the correct stochiometric ratio/area under the curve for H+ production in a given flow. Because Illumina only incorporates one nucleotide at a time and has a reversible terminator step, I get very reliable homopolymer length measurements. For Ion torrent, my error rate for homopolymer length measurement is about 1/100 for 3-4 nucleotids, and 1/10 for 5-6 nucleotides, and anything beyond that I give up on. For Illumina, the sky is pretty much the limit in my experience (we regularly work with deca-mer and greater lengths with high precision and accuracy). So, on this front, Mi seq definitely wins, and this can be important for copy number variation in the genome, allele-specific expression analysis in the transcriptome, and so-forth.


My personal assessment and uses of either instrument: We have chosen to use the Ion torrent for initial development of our panels. It's cheaper on a per run basis (which is key during R&D phase), and provides quicker turn around to bring new panels on line. We are leaning toward Ion Torrent our Acute panels (infectious disease, etc). We are leaning toward Miseq and Hiseq once our assays are validated on the Ion Torrent. The Miseq and Ion torrent costs are pretty much a wash when it boils down to useable data. However, when time is clearly not a factor and when we have the opportunity to batch/barcode/index multiple patient samples (with appropriate controls of course), the Hiseq decreases our sequencing costs ~10-fold, and thus increases our profit margin. If homopolymer regions does not affect your assay design, either Ion Torrent or Miseq is the way to go. If homopolymers are a problem, go with Illumina.

And thus, I will end by saying. If life tech (now Thermofisher) was moving forward with their Avalanche (code named Wildfire?) technology, I would be leaning more toward Ion Torrent for many applications. But I consider it a rather disappointing maneuver on their part not to push into that realm.

I think both platforms are quite good with the right pipeline for library preparations, which we've worked out :-) .

-Tom

[IonTorrent]

Hi there,

The information you heard at AACR is incorrect. As previously announced, the Avalanche chemistry is on track to be launched in 2013

[thomasblomquist]

hi there,

the information you heard at aacr is incorrect. As previously announced, the avalanche chemistry is on track to be launched in 2013

yay!!! :-)

[Elcannibal]

Hi there,

The information you heard at AACR is incorrect. As previously announced, the Avalanche chemistry is on track to be launched in 2013

Is that before or after the big boys review all the Life Tech programs ?

[cement_head]

What is "Avalanche" ? A link?

[thomasblomquist]

http://seqanswers.com/forums/showthread.php?t=23334

Avalanche is rumored to be Wildfire approach. (Key here is it is rumored). Essentially an answer to Illumina's bridge amplification approach.

This enables higher density reads per unit area, fewer heterodimer issues and polyclonal calls because it is initiated by ssDNA templates to create the sequencing "colonies." If Thermo could pull this off, they would have a serious contender to Illumina's platform on their hands. Homopolymers aside...

I have plenty of my own diagnostics that are not in homopolymer rich regions that could benefit from this approach to targeted sequencing.

-Tom

[Buzz0r]

...another rumor! Fabulous... ^_^

[NextGenSeq]

Avalanche is more than a rumor. LifeTech has shown actual data of on-chip amplification in their presentations.

[Buzz0r]

is that presentation online?

[ajthomas]

I was talking to a LifeTech guy (a technical specialist, not R&D, so take it for what it's worth) the other day and he mentioned Avalanche. As I'm still not clear on what it is exactly I asked him for more detail. While he didn't explain how it works, he did tell me that it will be used on the PIII chip and that he has seen perfect 600 bp reads obtained in house with it. If that's true, and if we actually see it at the customer level, that will really be something.

*Of course, this is all heresay at this point, so take it for what it's worth, which might not be much. We've heard similar grand claims from ONT and have yet to actually see anything available to us. It's nice to dream.