454 vs. Solexa to improve sequence quality and depth

[ebeyne]

Hello,

we study the genetic polymorphism (at transcriptomic level) between two species.
We are looking for some informations to improve our transcriptomic analysis.
For the moment, we had sequenced the transcriptome of two insect species by an 454-titanium half-run.

We would like to improve the sequence quality by either another 454 sequencing half-run either a Solexa sequencing (100bp, 15-20X coverage, more expensive)?

So, is anyone in the same case (454 done then choice between 454 and Solexa)? What have you chosen? Are the results (better depth with 454 or complement methods 454+Solexa) as good as you hoped?
Thanks for your help!

[NextGenSeq]

454 was the first to market and thus the oldest technology. It's also the most expensive.
Illumina is the current market leader. I personally wouldn't even consider purchasing a 454 system.

[ECO]

454 was the first to market and thus the oldest technology. It's also the most expensive.
Illumina is the current market leader. I personally wouldn't even consider purchasing a 454 system.

Age of technology has nothing to do with utility of the reads. I think you're letting your crotchety-ness influence your posting.

To the OP, at some point any read/mate length will saturate from an information perspective, where it saturates is going to depend on the characteristics of your genome and the read/mate lengths (and various other factors...seq methodology, library bias, etc). Given the same costs (and without knowing the size/repeat content of your genome, or if you have a reference already...) I would guess that mixing data types would get you to a more information-rich state.

[krobison]

Different technologies have different dominant error modes; mixing techs should enable better data.

Illumina will also allow much greater sampling depth, which would seem to be an overall win for any transcriptome project (lower expressed messages, more sampling of possible allele-specific effects, better precision in estimating expression level due to more counts). There are several papers in the cancer genomics field using mixture of 454 & Illumina with 454 providing scaffolding & Illumina piling on the read counts. IMO, the best summary of these can be found in a recent review.

[BaCh]

So, is anyone in the same case (454 done then choice between 454 and Solexa)? What have you chosen? Are the results (better depth with 454 or complement methods 454+Solexa) as good as you hoped?

Not exactly the same case: I'm more or less regularly doing 454 genomes of bacteria. But perhaps it might help you when I say that nowadays I *always* add in a round of Solexa to get rid of frequent errors in 454 (homopolymers, base switches etc.).

As I see it, neither of the technologies is superior to the other, they both have their strengths and weaknesses. Using both to cancel out the respective weaknesses has proven to be a very effective method for me. Just to give you an idea: my last benchmarks were: 2 errors in a 454/Solexa hybrid of a 4.something Mbp bacterium. I think that's not too bad.

454 was the first to market and thus the oldest technology. It's also the most expensive.
Illumina is the current market leader. I personally wouldn't even consider purchasing a 454 system.

I first wanted to answer with the "wheel" being a first to market and ancient technology, but let's keep it with sequencing: Sanger sequencing is from the stone age of sequencing. Yet, if I could choose between a well made Sanger paired-library and a well made 454 or Solexa paired library ... for paired-end, almost everything but costs (and perhaps problem with certain sequences) let the Sanger library win hands down.

Though I'm curious how PacBio might re-shuffle the deck there.

B

PS: But at least I learned a new expression: crotchety-ness. How lovely

[flxlex]

Not exactly the same case: I'm more or less regularly doing 454 genomes of bacteria. But perhaps it might help you when I say that nowadays I *always* add in a round of Solexa to get rid of frequent errors in 454 (homopolymers, base switches etc.).

Just out of curiosity, how do you do the actual error correcting, using what program/pipeline?

[BaCh]

Just out of curiosity, how do you do the actual error correcting, using what program/pipeline?

MIRA. Disclaimer: I'm the author of that beast, so I am probably not as impartial as I perhaps should.

That being said ... the easiest approach is 454 de-novo (or 454 + Sanger hynrid de-novo), then map Solexa against the whole multiple alignment (not only the consensus). MIRA flags each position where 454 (or 454 + Sanger) and Solexa disagree. Using a good finishing editor (I use gap4) it doesn't take long to go through all positions.

I would use 454 + Solexa hybrid de-novo only in the following cases:
1) long Solexa paired-end library (3kb)
or
2) low coverage of 454. Then combined with Solexa pe-lib.

Oh, and be wary of eventual GC bias in newer Solexa data. I have a data set of a bacterium (60%GC) with a few stretches at 45% and reads occurring there are ~3x more abundant than in the 60% areas. Same thing for a 45% bacterium with 35% areas. This breaks a lot of assumptions of de-novo assemblers.

B.

[ebeyne]

Hello,

thank for all yours comments about my problem! I add some details about my data.

Because of some technical problems, we must use the DNA library used to the 454 sequencing : it has the 454 adapters on both sides. So, to avoid to sequence only adapters, the company proposes us to concat the DNA library (5' and 3' joined) then to nebulaze the new fragment following a Illumina sequencing with the 100-pb kit (single-reads).

Does anyone know if the presence of secondary/tertiary structures in DNA (due to many adapter sequences) could prevent the nebulization to be random?