What Are You Using For A Reference Assembler?

Question

[diltsjeri]

I need some information on reference assemblers. What are you using? Which is the most preferable reference assembler?

[Neilfws]

What is a "reference assembler"? One that uses a reference genome, or one that you want to use as a reference for comparison with others?

[Neilfws]

Also: similar question, same user: http://www.biostars.org/post/show/44956/ion-torrent-reference-assembly/. Best to avoid posting multiple, highly-similar questions.

Answer 1

[Lee Katz]

AMOScmp-shortreads is working well for me, but it takes a bit longer.

[diltsjeri]

What files are you using to do the message file conversion with toAmos?

[Lee Katz]

I have been using toAmos_new (in the new versions only) to convert the fastq to a bnk, and then I start on step 20 in amos using -s 20 so that I can trick it into starting on a bnk file instead of an afg file. It's buried in my script but I think it's something like

toAmos_new -Q run.fastq -t SANGER -b amos.bnk
AMOScmp-shortreads -s 20 amos

You'll need amos.1con and amos.bnk in the same directory for this to work. You can use "amos" or any other prefix, but it must be the same between files.

[diltsjeri]

What's the advantage to starting with a bnk file? Also, are the options you posted above new to toAmosnew? I'm starting a pipeline with ion torrent data, so all I have is an sff. I use sffextract, to get the fasta,qual, and xml and I was going to use toAmos to convert to afg, but should I not?

[Lee Katz]

The only advantage is that toAmos_new can read in a fastq file and therefore you skip 1) converting to fasta/qual and then 2) converting to afg. Internally, AMOScmp converts first to a bnk anyway and doesn't use the afg anymore.

[diltsjeri]

thanks! this is really helpful.

[Lee Katz]

After validating it, AMOScmp unfortunately does not perform as well as I thought it should. I had a few more false-positives than when I worked with bowtie2. Sorry to do this, but I withdraw this recommendation in favor of newer tools. BWA came in as a close second to bowtie2 and was still better than AMOScmp.

edit I mean AMOScmp-shortReads.

[diltsjeri]

Thanks for following up.

[diltsjeri]

Bowtie doesn't output contigs though correct? I need my reads to be assembled.

[Lee Katz]

You'll have to follow up with samtools and "vcfutils.pl vcf2fq"

I am writing a script to automate this step but it is not finalized yet.

[diltsjeri]

My reference sequence (to be indexed) has ambiguous nucleotides. This is apparently not supported by bowtie. Have you ran into this problem and if so how did you work it? I can't just replace those V, H,etc with a nucleotide because it would make the alignment bias.

I noticed bowtie offers a -ntoa option on bowtie-build, but that just changes all N's to As. Wouldn't that create a bias? Also I have other nucleotide variables like V, H as stated above, which the option --ntoa wouldn't fix.

And I have some gaps :(

[swbarnes2]

replace all the ambiguous letters with N. You can do that with sed, or in a text editor like vim. And I'd use sed to get rid of - as well. Putting in an A will create a bias.

bwa will not crash on a genome like that. I'm pretty sure it will treat them all like N's.

[diltsjeri]

Bowtie won't take Ns, I wish it did ;(

Answer 2

[Nikolay Vyahhi]

• Bowtie
• BWA

[swbarnes2]

Those are aligners. Assemblers are like programs like vevlet.

[Nikolay Vyahhi]

Velvet is de novo assembler. If you need to assemble by reference, then you need aligner.

[diltsjeri]

This seems to be an ongoing debate on this forum.

[diltsjeri]

I need my reads to be assembled based on a reference. With tools like Bowtie and BWA I get the percentage aligned and I can see the aligned regions, but the reads are not being assembled based on the reference. I believe this is the difference between the two.

[Nikolay Vyahhi]

After alignment, you can construct (assemble) consensus sequence from BAM/SAM-file using samtools: http://samtools.sourceforge.net/cns0.shtml

[deepti1rao]

How can we have the snps and Indel replaced, and have the uncovered regions of the genome, represented as a series of Ns in the consensus assembly?

[ebaldwin]

I really wanna know the answer to this question.