[mira_talk] Re: Denovo versus mappings assembly with Mira
- From: Bastien Chevreux <bach@xxxxxxxxxxxx>
- To: mira_talk@xxxxxxxxxxxxx
- Date: Sat, 11 May 2013 23:22:16 +0200
On May 11, 2013, at 23:02 , David Coil <coil.david@xxxxxxxxx> wrote:
> But I'm wondering if the results of a mapping assembly (using the
> -SB:abnc-yes flag) count as a "genome sequence".
*sigh* The -SB:abnc flag. Combines the worst and the best of the worlds of
de-novo and mapping assemblies. Looks like way more people are using it than
I'd initially thought, maybe I should reconsider its removal in the 3.9 line.
> Basically you get one big contig (the mapping) and then a number of small
> ones assembled from the leftovers. But the big mapped contig has gaps even
> though it's called a "contig". You'd have to break the contig on those gaps
> to say, submit the genome to NCBI even though many of them are very small.
> One could of course use the unpadded result, but then you may be joining
> things together over large gaps that really should be considered separate
> contigs.
Careful there. Uncovered areas of a backbone (reference sequence) are NOT
deleted from the "unpadded" results, only "gap" columns are (i.e., columns
created by a spurious insertion base in one (or very few) reads at a given
place.
If you take the current default FASTA output from a mapping assembly in MIRA
you get something many people do not expect: an amalgam of the data from your
mapped strain and, in coverage holes, the data from the reference. I thought
this to be a good idea, but I'm not so sure anymore.
What one should do with the results from mapping: use convert_project to
extract the clean "by strain" data. Like this:
convert_project -f maf -t fasta mira_out.maf mynewresults
Uncovered areas of the backbone are then represented by a string of "N"
characters in these new results.
> So it seems to me that mapping assemblies are great for answering biological
> questions... but insufficient to say publish the genome sequence of an
> isolate. Would people agree or disagree with that idea? Am I thinking
> about the mapping assemblies incorrectly?
Agree, you are thinking about these assemblies totally correctly. I wrote the
mapping modes of MIRA to answer biological questions and that's what it does :-)
The following is just my 2 cents, other feedback welcome.
For *very* related strains however one can think reworking the mapping output
slightly in an assembly editor (gap4/gap5) and then use this for publishing.
Very related means in this case: whatever you feel comfortable with to finish
by hand … and on the importance you attach to having the genome "completed."
For strains having just a couple of SNPs, short indels and maybe two or three
larger indels or genome reorganisation breakpoints its a no-brainer, for more
you have to decide.
Hope that helps,
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:
