[amascarell]

Real Comparison

I hardly tried to find some papers comparing that technologies but I have not been successful. Do someone knows some review, or paper showing the best characteristics of all these machines (IonTorrent PGM, Ilumina miSeq and Roche 4545Junior). As far as I see, here all the discussions are about what providers say in the catalogue?
Thank you very much

[mucku]

Look at real data before you state it doesn't have a homopolymer problem. It really, truly does -- it isn't something that can't be overcome, but the instrument+software has to date had difficulty enumerating homopolymers, even ones shorter than 6.

Sorry if it came across that I stated it doesn't have any problems... I have to apologize.
At this point there is so much competition in that area that probably only half of the information actually fits the systems and the other half are optimistic forward looking statements...
As I said I hope it is not all marketing...

[krobison]

Look at real data before you state it doesn't have a homopolymer problem. It really, truly does -- it isn't something that can't be overcome, but the instrument+software has to date had difficulty enumerating homopolymers, even ones shorter than 6.

[mucku]

I found these webinars of IonTorrent very informative.

They also show data from the Broad Institute and others...

http://kevin-gattaca.blogspot.com/2011/06/introduction-to-ion-torrent-training.html

As I remember the pH sensor doesn't really have a homopolymer problem... I think a hexamer is still called with a Phred 20 score. So I think part of the actual error might be not from the chemistry but from polymerase slippage during library preparation... Also the quality at position 100 is still extremely good, and you probably don't have to cut back so much of your reads because of low quality ends.
The PGM has really potential, I hope that not all of it is marketing...

[Chipper]

PGM ~50 Mb 314 chip and 170 Mb 316 chip @ Broad, data presented at ESHG. ~1.8 M reads with Q > 17 if I am not mistaken.

[razibus]

Thank you Mucku for this relevant answer.

I did not notice my mistake and you are absolutely right : total output of GS Junior is 40Mb (specification from Roche) and it will be 80Mb at the beginning of 2012.

I agree with you, PGM looks good on the paper but we need more comparative scientific papers to have a better view of advantages/drawbacks of each machine.

[mucku]

- PGM : 10Mb (314 chip), 100Mb (316 chip), 1Gb (318 chip)
- MiSeq : 680Mb (2X100), 1Gb (2X150)
- GS Junior : 1Gb, 500bp max by amplicon

I have to correct your GS Junior statement:

The instrument produces on average 100.000 reads and 35 Mb per run. But I heard if you really have an extremely good library you can push it to almost 50 Mb... I got with 80k reads 38 Mb out once...

But I think that is the maximum of the GS Junior. The longer read lengths announced for 454 sequencing (700-1000) will only be available with a hardware upgrade to the big version.

I think you have to be careful when setting "read length" == "total output". Especially for de novo the read length is very important as you will need more coverage with shorter reads (e.g. Sanger =>7X, 454 =>20X, Illumina =>30X). Increases in read length will shift the numbers left... I am also not sure if you can really solve all problems of de novo assemblies using mate pair/paired end sequencing of short reads...
But for standard stuff like amplicons or short read mapping or experiments where you really need coverage you should go for PGM or MiSeq... probably PGM...

[razibus]

As stated by Mr Robison, because of primers and specific key sequences your inserts can't be longer than 150bp.

Ion 316 chip is just available to order in Europe. Ion 318 Chip should be available in September and the read lenght associated with this chip is 300bp.

[aleferna]

"I've heard the runs times will be relatively the same for 200bp, and for 316/318."

I thought the 318 was already out? Does anybody has some data on the 318? Does it really give you 200bp?

[razibus]

I gave prices for europe only and as you know prices are always higher here...

About PGM I can't remember if the run time is changed with 318 chip. Ask in PGM forum.

Whatever sequencer you chose, keep in mind that :
- ion torrent is very young and even if it seems to have a huge potential a lot of optimisations and improvements need to be done
- GS Junior has a good support and a well-developped protocol ; also the read size of amplicon is higher than the other ones
- MiSeq : No one really knows if it's real or not (just kidding) but to be serious we need to wait for an availability on the market and true specifications (miniaturization is not so easy)

[krobison]

Nextera cannot be directly used with PGM. With PGM's long read lengths, it would certainly be possible to design a "cheater" primer set to convert a Nextera library to a PGM library, but you would read through the Nextera primer on each read (and thereby waste some read length).

Also, at this time PGM libraries cannot have inserts longer than 150 nt; if your Nextera library is bigger than that it wouldn't currently be convertable.

PGM $50k
Server ~$15k, can run 2 PGM's.

I've heard the runs times will be relatively the same for 200bp, and for 316/318.

[aleferna]

Finally:
you say PGM is $109k(78k€)? Thought it was $50k what happen, is this because of the ridiculous server spec? Can't you run it with a PC or use an existing server? (I've seen people try to make money selling you a huge server to align and store but we already have that)

Also it says PGM 2 hours times 0.1Gb, I guess this is for the 314?? For the 318 is it 2 hours for 1GB?

[aleferna]

Also we've been using the Nextera kit and we are delighted with it, anybody know's if you can use it for PGM?

[aleferna]

Thanks so much for the info, looks like the PGM has a huge lead on the other 2 (particularly since its a new tech, expecting exponential growth)... I wonder if the device quality/warranty also holds. Also we talked to an Illumina rep and he was short selling the GAIIx, I think they realized they can't compete with the miseq?