[maker-devel] tblastn Cleanup?

[Carson Holt]

There are several things.  I set several filtering options directly on the
BLAST command line.  These are things like maximum intron length, an e-value
filter, and simple repeat filtering (called dust filter in NCBI blast and
seg filter in WUBLAST).

I also run repeat masker over the genome first.  This allows simple and
complex repeats to be removed before running BLAST (otherwise you get many
false alignments).

Last I filter the results based on percent coverage of the hit to the
original database sequence and percent identity.  I think you can set
percent identity as a flag in BLAST, but the percent coverage filter is
being calculated by MAKER, so to do this outside of MAKER would require that
you write your own filtering script to compare the length of the alignment
to the length of the sequence in the database.

I also have an HSP depth overlap filter.  This removes weird low complexity
hits that escape repeatmasking.  They show up as multiple HSPs overlapping
multiple times in the same region (usually very high numbers like 90 HSPs
all 100 bp long in the same region).  I calculate the number of base pairs
in the alignment on the hit then divide by the number of base pairs in the
query alignment.  If it's greater than 3, I throw the hit out.

Thanks,
Carson



From:  Yogesh <yogeshp08 at gmail.com>
Date:  Tuesday, 15 May, 2012 12:07 PM
To:  <maker-devel at yandell-lab.org>
Subject:  [maker-devel] tblastn Cleanup?

 
Hello,

I have a few tblastn alignments with a lot of low quality hits. I have to
clean that up. Can you please suggest how Maker pipeline does it? Also can I
run it directly on my data without having to go through the whole pipeline?

Thanks,

-Yogesh
_______________________________________________ maker-devel mailing list
maker-devel at box290.bluehost.com
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20120518/8e68fe59/attachment-0003.html>