Assembling Illumina mate pairs using ABySS

[arrigon]

Dear all,

I plan using ABySS for assemblling a set of paired-end reads along
with a library of mate pairs. As you know, the paired-end libraries
have read pairs that are facing each other (forward-reverse
orientation), whereas mate-pairs are oriented forward-forward.

Thus, do I need to reverse-complement one of the reads set of my mate
pair library, to make it consistent with a forward-reverse
orientation, prior to assembly with ABySS?

Thanks in advance for any information; its seems that this point
remains somewhat ambiguous on the forums and I would appreciate having
a clear answer for that ;).

Nils

[Shaun Jackman]

Hi Nils,

No, you do not need to reverse complement the mate-pair reads. ABySS
determines the orientation automatically. However, I recommend using the
mate-pair library for scaffolding and not for sequence data by
specifying the paired-end library using the lib parameter and the
mate-pair library using the mp parameter, like so:
abyss-pe lib=library1 mp=library2 library1='foo.fq' library2='bar.fq'

The names library1 and library2 are arbitrary. You can name the
libraries however you like.

Cheers,
Shaun

[arrigon]

Hi Shaun,

Thanks for this answer!

Best,
Nils

[T.Hamdi Kitapci]

Hi Shaun,
Why do you recommend using mate-pair for scaffolding only? Rather than using for both contigs and scaffolding ?

Thanks
Best Regards
T. Hamdi Kitapci

[T.Hamdi Kitapci]

Hi Shaun,
Why do you recommend using mate-pair for scaffolding only? Rather than using for both contigs and scaffolding ?

[Anthony Raymond]

Hi Hamdi,

We usually don't use mate-pair (MPET) data for anything but scaffolding because MPET data contains chimeric reads created by sequencing through biotin. This means the MPET sequences will add a lot of erroneous kmers to the assembly if used in the debruijn graph.

When it comes to contiging, the issue is that MPET data contains both FR (typical for PET) and RF (desired in MPET data) oriented reads. The scaffolding stage is built to handle this complexity, but the contig stage is not.

If you'd like further reading, take a look at this: 

http://products.illumina.com/content/dam/illumina-marketing/documents/products/technotes/technote_nextera_matepair_data_processing.pdf

In particular, figure 2 show's the issue I'm referring to.

Cheers,

Tony

--
You received this message because you are subscribed to the Google Groups "ABySS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to abyss-users...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Albert Vilella]

Have you tried using https://github.com/sequencing/NxTrim on beforehand?

A.

[Anthony Raymond]

Hi Albert,

I have not, but that would work nicely to split the data into the different categories. You'd still want to use the "unknown" and "mp" reads in scaffolding only. With this you could probably use everything but the "unknown" reads in the single end stage since the chimera's would be removed.

Thanks for posting the link :)

Cheers,
Tony