[mira_talk] Re: allow for more polymorphisms
- From: Oscar Franzén <oscar.franzen@xxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Tue, 08 Dec 2009 10:10:18 +0100
Bastien Chevreux wrote:
On Montag 07 Dezember 2009 Oscar Franzén wrote:
I'm trying to assemble a highly polymorphic genome using MIRA, but my
problem is that the final genome created by MIRA is almost twice the
haploid size of the genome (output size of all contigs is around 24Mbp
but the haploid genome size should be around 12 Mbp). I think that the
high polymorphism rate causes MIRA to duplicate all contigs. So my
question is, is there an option/parameter in MIRA to increase the
allowed polymorphism rate?
I'm using the latest version of MIRA and the data type is 454 and I'm
running a 'normal' assembly with default parameters.
I suspect that this falls under the no-can-do. Technically speaking, sequences
from different ploidies with base differences very much look like a repeat to
an assembler. In a perfect world, with even coverage everywhere, one could
distinguish them by looking at the coverage, but ... I dare to say that this
would be a pretty daring thing to try with actual data (be it Sanger, 454 or
Solexa).
The master switch to turn off recognition of repeat marker bases is -CO:mr,
setting it to no will stop it from working. But then a whole new range of
problem arises and you will need to tighten a few other parameters.
If you really want to try, these would be the parameters I'd change (untested,
you might want to test on a smaller data set first): in Smith-Waterman
alignments, crank up the minimum relative score (-AL:mrs) to 90 or 95% for all
sequencing technologies used. In the contig parameters, switch off the marking
of repeats (-CO:mr=no) but at the same time you absolutely must make the
contig assembly more strict: decrease -CO:rodirs to 10 (or perhaps even some
value between 5 and 10) for all sequencing technologies used.
I think you will still run into some troubles with repeats, but it's worth a
try.
Regards,
Bastien
Thanks! It's definitely worth a try.
/Oscar
--
You have received this mail because you are subscribed to the mira_talk mailing
list. For information on how to subscribe or unsubscribe, please visit
http://www.chevreux.org/mira_mailinglists.html
Other related posts:
