Assembly for both 454 and GAIIx

[rtp115]

Hello,

Could anyone recommend a good assembly program that can combine both 454 and GAIIx reads? I'm hoping to build better scaffolds than individual assembly.

Thanks in advance!

[natstreet]

There was a brief discussion about this on the ABySS mailing list recently. From memory the general advice was to try something like a de novo of the Illumina data and to then chop up the contigs and feed them into Newbler along with your 454 data.

The approach will also be dependent on your genome size. For example, if the genome is not big you could try MIRA.

I think clc (commericial so maybe not an option) can handle combined assemblies.

Finally velvet has the option to include long read alongside short reads datasets but your mileage with Velvet will be dependant on the amount of data you have and your computing resource.

[themerlin]

I would recommend Mira. I wouldn't try a true denovo assembly though. There's a lot of good info on the Mira website.

[Torst]

From memory the general advice was to try something like a de novo of the Illumina data and to then chop up the contigs and feed them into Newbler along with your 454 data.

Newbler (gsAssembler) can only accept reads up to 1024bp. I have tried splitting it into fake PAIRED reads, but it didn't help much. I guess the 3kbp insert of the 454 reads trumped the 200bp insert of the Illumina PE reads.

The approach will also be dependent on your genome size. For example, if the genome is not big you could try MIRA.

Worth a try, I haven't investigated this option enough yet.

I think clc (commericial so maybe not an option) can handle combined assemblies.

Yes it can, but its results are variable in quality.

Finally velvet has the option to include long read alongside short reads datasets but your mileage with Velvet will be dependant on the amount of data you have and your computing resource.

Adding 454 reads as as -long tends to make things worse in my experience. It is unclear why this is the case, it is possibly homopolymer errors in the reads. Even shredding 454 scaffolds into long read PAIRS didn't help much.

Ultimately we use the Illumina data to fix homopolymer errors in the 454 scaffolds (assuming it was mate pair 454) and found not much to gain from Illumina PE data. But that may change when Illumina MP data (3,5,10 kbp insert) is clean and reliable.