how to obtain assembled fasta sequence with velvet?

[foxyg]

I am trying to do some denovo assembly using velvet. First of all, is Velvet the best or one of the best devono assemblers available?

I followed the manual I used velveth and velvetg to obtain all the graph file and contigs.fa but I don't know how to obtain a fasta file for each sample. Does anyone know how to do this?

Thanks

[maubp]

Um... contigs.fa is a FASTA file. What do you mean by each sample?

[westerman]

Perhaps 'foxyg' wants an individual file for each of the contigs (aka, nodes) contained within the contig.fa? Such a tool should, of course, be in the 'quick access' part of anyone's toolbox.

[foxyg]

Um... contigs.fa is a FASTA file. What do you mean by each sample?

I meant for a consensus fasta sequence.

[westerman]

A consensus sequence out of a denovo assembly without any mapping information? Might as well just throw in a bunch of 'X's (unknowns) in place of the fasta headers. E.g.,

sed -e 's/>.*/XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX/' file.name

But I suspect that this won't be want you want. Or maybe it will since I can think of some projects that want a single sequence in their FastA.

I think you'll have to give a bit more complete information on what you have and where you want to go with it for us to give you much more help.

[foxyg]

I have illumina reads of 1 sample on chr 17 only, I want to do a denovo assembly to compare the consensus sequence from denovo to the one from alignment.

[westerman]

I have illumina reads of 1 sample on chr 17 only, I want to do a denovo assembly to compare the consensus sequence from denovo to the one from alignment.

Can't be done. Without other information it is not possible to stitch the contigs together in a meaningful fashion.

It would be useful to know why you want to compare a de-novo assembly against a known reference. Most people would more simply and usefully map the reads against the reference without doing an assembly -- you can get a lot of the variant information from doing this.

However there are some possible reasons why you might wish to do the assembly and then comparison. I can think of two off the top of my head.

1) You are doing some sort of test of "how well does a de-novo assembly work".

2) You wish to find potential insertions in your sample; e.g., reads that form a good assembly which lack a strong corresponding map into the reference.


What I would do is to take the contigs (not strung together) and blast or blat them against your reference sequence. From that you can either

1) Figure out how well your denovo assembly worked (e.g., from the matched sites)

or

2) Figure out which contigs have end(s) in the reference but also a large number of bases not in the reference; these will be your potential insertions.

Of course you may have a entirely different goal for your project. But in any case without further information there is no way to string a bunch of unrelated contigs into a coherent single contig.

[foxyg]

Thanks, I am interested in some structure variance that might not show up in regular alignment. I think the best approach is to use the contigs to blast against reference like you suggested.