Improving de novo assembly

[Anelda]

Hi there,

I am assembling 454 bacterial genome reads using GS Assembler/Mapper. There is no complete reference genome out for my bacteria - only a draft i.e contigs/scaffolds. I was thinking it may be useful to first do a mapping to the draft contigs using all my reads. Then doing a de novo assembly using all my reads. Then doing an assembly of the resultant fasta contigs of the two processes.

Does anyone have an opinion about this method? Is there software that I could do the second step with?

Thanks!
Anelda

[maubp]

You could also try other 454 assemblers, e.g. MIRA3

[wwq413]

Mapping can get good SNPs, but I don't think mapping can find divergent region or big INDELs.

[sophied]

These guys have done a similar thing to what you are suggesting, Anelda, although they used Illumina reads rather than 454. It might be an interesting read...

http://www.pnas.org/content/early/20...39108.abstract

[dsenalik]

Hidden deep in the gsAssembler output files is information about what contigs connect to other contigs. So even thouugh it looks like a pile of contigs, there is a limited number of ways they can go together. So it may not be as hard to de-novo assemble it as you think. (if you are lucky)

See Lex Nederbragt's excellent description of the 454ContigGraph.txt file for more information at http://contig.wordpress.com/tag/454contiggraph-txt/

I wrote a hard-to-use program to visualize this information, if you are interested it is at

http://www.vcru.wisc.edu/simonlab/sdata/software/index.html#contignet