[maker-devel] model_gff not in output

[Carson Holt]

The only way MAKER should ignore a legacy annotation (only what's in
model_gff is considered legacy by MAKER) is if, you also set one of the ab
iniio predictors to run simultaneously or provide a pred_gff file and one
of those models scores higher and would overlap the legacy model.  Then
MAKER chooses that model instead.  Also if you have two overlapping legacy
models MAKER will only keep one or the other under these same conditions.
MAKER will only keep all legacy models regardless of overlap if you only
supply model_gff with no other predictors turned on.  Once you turn a
predictor on, MAKER takes this as a cue that you are letting it make
changes.  Legacy models should always be kept under all circumstances if
there is nothing overlapping them with a higher score.

Are the missing models partially overlapped by anything in the resulting
MAKER annotations?

Thanks,
Carson



On 12-09-04 4:59 AM, "Michael Thon" <mike.thon at gmail.com> wrote:

>I'm using maker to update a legacy annotation.  As input I'm using
>RNA-Seq aligned with cufflinks, ESTs, provided in fasta format, and
>proteins downloaded from UniProt.SwissProt.  I have done two runs of
>maker so far.  The first one using the legacy annotations in both the
>model_gff and pred_gff parameters.  In the second run I used the legacy
>annotations in model_gff and in pred_gff I included gene models created
>with GeneMark-ES.
>
>In both runs 1 and 2 I have found two genes (so far) that exist in the
>legacy annotations but are not in the final gene models output by maker.
>Both genes have overlapping cufflinks annotations, in addition to having
>annotations in model_gff. I thought maker was supposed to keep all the
>annotations in model_gff, only replacing ones in which it could find an
>alternative model with better support.  Is there any case in which is
>will remove a model?
>
>
>Another discrepency I found in run1 is a gene that maker 'moved' upstream
>approx. 150 bases.  The gene locus annotated by maker covers the original
>annotation, but the CDS does not. The site of the original CDS is covered
>by an annotation in model_gff, pred_gff, two ESTs and a cufflinks
>annotation.  Maker still seems to have moved is is upstream where it only
>has an overlapping cufflinks annotation.  the three-prime utr annotated
>by cufflinks still covers the legacy annotation though.
>
>Here's a link to download the maker gff file I'm looking at:
>
>https://dl.dropbox.com/u/320712/supercont1%252E1.gff.zip
>
>The genes that are in the legacy annotation but missing in the maker
>annotation are:
>GLRG_00074 and GLRG_00092
>
>the 'moved' gene model I described is model GLRG_00081.  they all within
>the first 350 K of sequence.
>mike
>
>
>
>
>
>_______________________________________________
>maker-devel mailing list
>maker-devel at box290.bluehost.com
>http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org