[mira_talk] Re: prescreening illumina sequences
- From: Bastien Chevreux <bach@xxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Tue, 1 Dec 2009 20:47:06 +0100
On Dienstag 01 Dezember 2009 Reith, Michael wrote:
> Thanks for the comments Bastien. It turns out that I have 36% N's (from
> paired end, 76 bp reads from a lower eukaryote (genome ~35 Mb)). Two
> questions:
>
> 1. Since filling up the computer memory with a ton of N's seems like a
> waste, I'm going to remove all sequences that don't have at least 10
> bases of Phred > 15.
I'd suggest not to do that in a first time. The quality values written by the
GERALD pipeline are, well, not the most trustworthy ones. I've seen a lot of
cases of perfectly valid sequence with low quality values and, vice versa,
sequence with good quality values to be entirely unusable.
> If I also chop off the runs of N's at the 3' end
> the remaining sequences,
Yes, chop off long runs of N's. I might even implement such a clipping in MIRA
one day, but not before releasing 3.0.
> is Mira OK with having different length Solexa
> sequences
MIRA does not care about different lengths :-)
> , or should I just let Mira do the chopping?
The "memory saving chopping" is not existing. -CL:pec will handle things well
from a clipping point of view, but the sequence will just be hidden and thus
eat away memory even if not used.
> 2. Should I be complaining to my sequencing service? Has anyone had
> experience with these read lengths from paired ends?
I would. Here are some numbers I saw in the last few months.
- in the very worst project (paired-end, 36bp, 68% GC), MIRA had to clip away
a total of 17.5% of all the bases, killing 1 milliion reads completely out
of 8 million (12.5%). I don't have separate numbers for 'N', but these are
generally only a tiny fraction of the clipped sequence.
- in 'normal' projects (76bp, quite neutral GC), MIRA clips away between 8 and
12% of all the bases.
Note that % GC has an influence on the sequencing quality, with higher GC
being potentially quite detrimental to sequence and quality values. See also
the GGCxG problem I describe in
http://chevreux.org/GGCxG_problem.html
The clipping works really, really well. After that, 80% of the (possibly
clipped) Solexa 76bp reads contain no error, 10% one error, 5% two errors, ~2%
three errors and the rest more errors.
> Is this typical or
> indicative of problems with the sequencing and/or base calling?
Talk to your provider, asking them to put the number of N's (36%, sheeeesh) in
relation to other projects they delivered. If they tell you "that's normal"
with a straight face, ask around other sequencing providers about their
numbers :-)
Regards,
Bastien
--
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:
