Templated assembly with velvet columbus module

Question

[susannehoward]

I have used the columbus extension of velvet to do a templated assembly, which, as far as I can tell, was successful. Until someone pointed out to me that he would not expect to see numerous contigs piling up in the same region of the ref seq? (I used IGV for the visualization). He said he would have expected to see only one contig, these numerous contigs having been merged? Has anyone used velvet columbus and got output other than contigs.fa?

the distribution of contigs looks remarkably similar to a denovo assembly and I guess the idea behind his question might be why the contigs are not merged since the reference genome should have provided a means of ordering the contigs.

Any thoughts?

Susanne

Answer 1

[Charles Warden]

I'm not sure about the results with all the various types of parameter configurations.

For example, I believe VelvetOptimiser is supposed to be able to help with some of that, but I don't know how those work with the Columbus module (or whether you have to start with running regular Velvet).

That said, if you think you have an expected haploid genome and a fairly good starting sequence (with only SNPs and small indels), I would consider using the VarScan consensus command (and convert the "Cons" column to a FASTA sequence).

For example, here is one command that I used (as an alternative to a de-novo assembly):

java -Xmx8g -jar /opt/varscan/VarScan.v2.4.2.jar pileup2cns [input.pileup] --min-avg-qual 20 --p-value 0.05 --min-reads2 4 --min-coverage 10 --min-var-freq 0.3 --strand-filter 1 > consensus_table.txt

For some reason, I think there was an advantage to using the slightly older version of VarScan (v.2.4.2), but I don't know how much of a difference that makes.