[mira_talk] Re: Assembly in versions 4_05 is worse than previous versions
- From: David Coil <coil.david@xxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Thu, 2 Jan 2014 07:52:24 -0800
Thanks for that Bob! Good to hear some hard data.
David
On Wed, Jan 1, 2014 at 8:34 PM, Robert Bruccoleri <
bruc@xxxxxxxxxxxxxxxxxxxxx> wrote:
> Dear David,
> I have hard data on an artifact with Nextera -- it cuts with
> non-uniform probability. I have analyzed sequencing data from the same
> bacterial DNA prepared twice with Covaris fragmentation and once with
> Nextera, where I counted the number of repeated 50 bp read fragments (all
> taken from the 5' end of the actual reads), and the number of repeated
> fragments is an order of magnitude larger with Nextera than ultrasonic
> fragmentation.
> I've also analyzed the frequency of cuts in an 16s rRNA amplicon using
> Nextera and Covaris fragmentation, and the Nextera cuts are wildly uneven
> (ratios over a 100 are seen), whereas the Covaris cuts are underrepresented
> near the ends of the amplicon, but are otherwise relatively even (biggest
> ratio is 6).
> Unfortunately, the bacteria in these studies are proprietary to one of
> my customers, so I can't share the data publicly, but you can easily repeat
> the experiments.
>
> Nextera works for human genome resequencing because variant calling
> only requires that a threshold of coverage be exceeded throughout the
> genome, and these cutting biases generally don't affect the coverage enough
> to affect achieving the threshold. In contrast, Mira generally depends upon
> the cut sites being uniformly random.
>
> Cheers,
> Bob
>
>
> David Coil wrote:
>
> Irrespective of the particular discussion at hand, I would love to see
> someone (Bastien?) expand on the following piece of it:
>
>
>> Second: if the Illumina sequencing was done with Nextera library prep,
>> things look pretty bleak: I’ve come to hate Nextera data sets as many of
>> them are dirty beyond imagination, riddled with all kinds of artefacts
>> which lead to severe problems not only in MIRA, but in many assemblers I
>> have on my list of programs to regularly look at.
>>
>>
> I have heard rumors to this effect before, but no hard data. I brought
> this up at a recent assembly workshop and the general feeling from folks
> there was "Nextera isn't so bad".
>
>
> David
>
>
>
>
Other related posts:
