[mira_talk] Re: Very long transcripts
- From: Martin Mokrejs <mmokrejs@xxxxxxxxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Wed, 31 Oct 2012 12:00:36 +0100
Hi Jackie,
you haven't said what lab protocol was used to prepare the sample for
sequencing.
The numbers sound like a whole disaster and therefore I guess you used the
MINT/SMART
protocols (especially as you mentioned polyA/T trimming and intended 3'-end
only cDNA
library). I analyzed few dozens of such datasets and FIXED them. Since I
started to
work on these I hit so many issues that honestly said, my only advise is you
don't
work with such datasets unless you clean them up. The question is what you
should
do ... ;-) I found the way through ... If you want to re-invent the wheel
reserve
6-8 months of your time as a molecular biologist and hire a programmer.
On the other hand, if you used Roche protocol then with those random hexamers
you
shouldn't have that many issues but still I have seen datasets which also had
polyA/T
tails in reads. These, in theory, should NOT appear in cDNA prepared by Roche
protocol
but somehow they do! ;-)
I think I mentioned that here on this list some while ago ... I can offer a
commercial service in cleanup and correction of the SFF files. There have to be
done
so many alignments that fixing 300k reads (a 1/4 XLR plate) takes several weeks
on a
fastest machine I could get). Therefore, a paid service. I really do know why it
takes so long ;-), not a single adapter-trimming software available around does
anything even remotely similar and unless the thing is published I won't say
more
about my approach.
In general, transcripts are 2-5kb long, some aberrant transcripts are longer
with
extended 3'-UTR. That's the current view of transcriptomes. Unless you have a
truly
obscure organism then the assembly is just wrong.
As a quick insight I can look into the assembled contigs for adapters and
..., but
for a final solution I would need SFF files. I can work with fast(a/q) files
but that
is suboptimal. So once again. How many raw reads are to be analyzed? What
protocol?
Otherwise? You have to wait and meanwhile, good luck with you efforts. ;-)
Best,
Martin
Bastien Chevreux wrote:
> On Oct 30, 2012, at 14:35 , Jackie Lighten <jackie.lighten@xxxxxx
> <mailto:jackie.lighten@xxxxxx>> wrote:
>> I have performed an accurate de novo assembly with poly-a/t trimming.
>> I get all reads assembled, and no singlets, into around 66k contigs. Around
>> 27k of these are large contigs, with the largest being ~25k bases long. This
>> does not make much sense to me as I constructed a 3' target cDNA library
>> (454 FLX). I can envisage multiple open reading frames may create longer
>> transcripts but 25k seems dodgy to me.
>> Any thoughts?
>
> Yes. Have a look at those contigs :-) No joke, this always brings the best
> insights.
>
> Possible reasons:
> - PKS genes. These can be up to 45 - 50kb long, maybe even longer
> - contamination of the cDNA with gDNA
> - introns. Especially for highly expressed genes, one has a higher chance to
> have sequence unedited mRNA
> - unclipped adaptors which "join" contains
> - assembly "errors": short overlaps of just a couple of bases
>
> B.
>
--
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:
