[mira_talk] Re: Getting observed insert size etc for paired end reads
- From: Bastien Chevreux <bach@xxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Thu, 13 Feb 2014 11:32:32 +0100 (CET)
> On February 13, 2014 at 10:48 AM Peter Cock <p.j.a.cock@xxxxxxxxxxxxxx> wrote:
> [...]
> Those are the predictions for each of the five passes - settling down to a
> mean
> template size of approx 230, standard deviation 114, skew 0.77, min 24, max
> 480.
>
> However, the MAF header seems to have the min/max from the first pass (22,
> 515),
> is that an error?
Yes. And no, it currently depends a bit on your manifest file.
If the segment_placement is not known, MIRA determines the placement in the
first pass and the size in the second pass. If you gave the placement (other
than "unknown") as parameter in the manifest, then the size is directly
determined in the first pass. Though maybe I should change that as first pass
assemblies are a bit of a "warm-up" and can include quite some misassemblies.
That's more of a problem for large libraries I think, though I need to ponder
about this whether I should not push that to second pass per default.
Size predictions from subsequent passes however are ignored on purpose as MIRA
will have used the previous predictions to place reads and thus have rejected a
couple of pairs which fell outside the prediction. If one recalculates
predictions on that, the results will be ever shrinking ranges which is not what
you want.
B.
Other related posts:
