[mira_talk] Re: Multiple long repeats in genome.
- From: Andrzej N <andrzej.k.n@xxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Tue, 22 Mar 2011 13:37:33 -0500
> What you are describing only affects contigs made of repeats longer than a
> read length (or than the insert size of paired reads).
>
> In the past, MIRA indeed made multiple copies of repeats and stored them
> separately ... because I too feel that this is the right approach.
>
> Until people complained that those version of MIRA were bad program because
> they made "much more" contigs than other assemblers. When I then started to
> see assembler comparisons involving MIRA just based on N50 and number of
> contigs, I mailed the authors of some to point out the important difference.
> I either got no response, or responses from people I had then to assume were
> undergrad students who had no idea about the underlying problematic and had
> this as a "homework" or response from people with enough knowledge but told
> me that the "needed something easy to measure and they had no time for more
> in-depth analysis of the assembly quality".
>
> I then grudgingly reverted my decision and made MIRA again collapse
> repetitive contigs.
>
> Sad, but true.
>
Very sad :(. This also if affecting a mapped assemblies, leaving a "holes"
in some regions, imagine you have three region of long repeat (let say
3000bp), technically two of them will be "empty", on the other hand it's
quite easy to check which region have repeats... BUT when you build new
contigs and do reassembly using already build contigs... it's like running
after your own tail ;). I also know this is probably on problem if someone
don't have pair end reads...
PS. I don't even care any more about N50..., until I will not see contigs
:). This gives SOME info but...
> > I hope MIRA is not deleting that repeats. How to keep them together
> with my
>
> > other contigs?
>
> Erm, what do you mean by that?
>
> B.
>
In sense, that ALL contigs are keep by mira. When I will choose "best
contigs" I should get all that one which are build from repeats?
Andrzej
Other related posts:
