454 + Illumina Combined Assembly

[Kmart]

Hi All,

I am looking for suggestions on how to combine data from both 454 and Illumina platforms. Briefly, I have 454 sequenced BACs with average insert sizes of 150kb covered to approximately 30X and an illumina 100bp PE genome of the same species covered to ~25X. I am interested in obtaining the least gapped assembly of the BAC sequences only - not a completely assembled genome of this species.

The only assembler I have attempted is velvet with -long libraries in addition to the -shortPaired library and this does not produce an assembly of comparable quality to my 454-only newbler 2.5p1 assemblies of just the BAC sequences. Any advice is greatly appreciated.

Best,
Kyle

[nickloman]

I wrote a blog post on this subject:

http://pathogenomics.bham.ac.uk/blog/2011/08/assembling-illumina-and-454-data/

I'd be inclined to add your Illumina reads to a Newbler 2.6 assembly as your starting point.

[Kmart]

Thanks nick,

I had a look at your blog and I can see there are several alternatives. Just to clarify, when you say "add your illumina reads to a newbler 2.6 assembly" what program would you recommend for a first shot? Presumably not newbler itself, but a de Bruijn assembler like velvet?

Best,
Kyle

[nickloman]

I mean run Newbler with the 454 and Illumina reads. V2.6 supports Illumina FASTQ files and seems to do a pretty good job with the bacterial datasets I've tried it on. If you are finding it takes too long consider using the -large flag.

[Kmart]

Can newbler really manage the quantity of data produced by Illumina? I have 171 million 100bp PE reads. It was my understanding that the typical OLC assemblers could not handle this.

[Seqasaurus]

For what its worth, i can vouch for newbler 2.6 with 1.2 million titantium plus up to 2million HiSeq 100bp SE. did a good job with eukaryotic cDNA

[nickloman]

I am not sure whether Newbler incorporates the FASTQ reads as part of the overlapping stage or whether it uses some kind of mapping assembly algorithm, but I would certainly give it a try and see how it goes. Obviously there's always a danger of running into memory or CPU issues with that many reads.

[bioBob]

Because you are wanting only the BAC portion in your final assembly, have you thought of filtering the Illumina data for reads that have, say, 50% identity to the 454 reads prior to a combined assembly?

[v_kisand]

Nick has a good blog, I would go for MIRA3.

-veljo

[sphil]

MIRA3 has done really good jobs here, assembling yeast 454 and ill PE 90x90 data.

give it a try!