[maker-devel] clarification on creating a standard build

[Carson Holt]

You will get the ID from the IntrepProscan report. Then you can take that ID and grep for it in the MAKER gff3.

All ab initio predictions have match/match_part features created for them the gff3. You can then take those non-gene match/match_part features and provide them to pred_gff.

You then have two alternate ways to get those models into your dataset.

1. Do a second run with only that pred_gff (i.e. turn off all other MAKER options and blank out all evidence including repeat masking options) and set keep_preds=1.

That will simply take the pred_gff match/match_part values and turn them into a nicely formatted gene/mRNA/exon/CDS features together with associated fasta files. Those can then simply be merged into your current result using GFF3 merge.

2. Provide maker_gff (set pred_pass=0 and all other pass options to 1), provide pred_gff, and set keep_preds=1.

This is the same as the previous run option, but MAKER will do the merging for you. But it will take longer since it will use the maker_gff to rebuild all models and evidence in memory and rescore everything.

—Carson



> On Mar 20, 2018, at 6:48 PM, Valerie Soza <vsoza at uw.edu> wrote:
> 
> Hi MAKER community
> 
> I am trying to create a standard build as indicated in the Campbell et al. 2014 papers in Plant Physiology and Current Protocols in Bioinformatics. I was following the protocol as outlined in Current Protocols in Bioinformatics, but then came across this thread in the MAKER google forum: https://groups.google.com/forum/#!searchin/maker-devel/quality_filter%7Csort:date/maker-devel/97aNJkT3bgk/mpL7V5QWAAAJ.
> 
> I can’t reply to this original thread, but I am trying to follow Carson’s suggestion for a standard build using this protocol instead now:
> "One note I’d like to make, is that doing a second round with keep_preds=1 is the wrong procedure (only do that if you really want to keep everything - i.e. in some fungi or oomycetes). Rather you should use InterProScan to evaluate the rejected models in the non-overlapping.abinit.proteins.fasta file, then grep the ones that have an IPR domain out of the GFF3 (will be match/match_part features) and then pass them to pred_gff in a separate run (just updates the format to gene/mRNA/exon/CDSwith proper reading frame). You can then merge the resulting GFF3's and fasta files.”
> 
> Instead of doing a second round of annotations with keep_preds=1, I am using my original annotations with keep_preds=0. I have used InterProScan on the non-overlapping.abinit.proteins.fasta. I am unclear as to what gff3 file to use to grep for genes with IPR domains from the non-overlapping.abinit.proteins.fasta file. Genes from the non-overlapping.abinit.proteins.fasta file are not in my .all.gff file created by the gff3_merge script. 
> 
> What gff3 file should I be using to resurrect proteins with IPR domains from the non-overlapping.abinit.proteins.fasta? Should I be doing an annotation with keep_preds=1 as well, and resurrecting genes with IPR domains from this gff3?
> 
> Thanks.
> 
> -Valerie
> 
> Valerie Soza, Ph.D.
> c/o Hall Lab
> Department of Biology
> University of Washington
> Johnson Hall 202A
> Box 351800
> Seattle, WA 98195-1800
> 206-543-6740
> http://staff.washington.edu/vsoza/
> 
> 
> _______________________________________________
> maker-devel mailing list
> maker-devel at box290.bluehost.com
> http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org