keep_preds=0 but some gene models have anyway an AED equal to 1.

classic Classic list List threaded Threaded
4 messages Options
Reply | Threaded
Open this post in threaded view
|

keep_preds=0 but some gene models have anyway an AED equal to 1.

Jacques Dainat-4
Dear all,

It is not the first time I see that, but I have an annotation launched with the option keeps_pred=0 that contains gene models with AED score equal to 1. As far as I understand, AED score to 1 means there is no evidence support. So it should be purely abinitio without any line of evidence in front of this prediction.
But as it is written in “Michael S-Cambell et al., Genome Annotation and Curation Using MAKER and MAKER-P,  Current Protocols in Bioinformatics, 2014” page 4.11.36 about the the keep_preds option:
"MAKER rejects models that do not have at least some form of evidence support.” Setting keep_preds to 1 “remove the evidence support requirement”.

So, how should I understand my results? Some predictions with low support (but support anyway) can have AED score equal to 1? Or some purely abinitio prediction without support at all can anyway be selected even when keep_preds=0 is set up? 

Best regards,

Jacques Dainat


_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
Reply | Threaded
Open this post in threaded view
|

Re: keep_preds=0 but some gene models have anyway an AED equal to 1.

Carson Holt-2
By default MAKER will not let models through without at least some degree of evidence overlap (AED < 1). This is because ab initio predictors overcall (sometimes by as much as a factor of 10, i.e. 10 false positives for every true positive). You can dial in the minimum AED support using the AED_threshold. But if yo8u want to keep everything that does not have overlap with a better supported model then setting keep_pred=1 will allow even unsupported models to be maintained (1 for yes and 0 for no). Situations where you may want to do this include passing in old annotation datasets or working on organisms with very high genedensity and low ab initio false positive rates (many species of fungi meet this criteria).

—Carson


On May 2, 2018, at 5:55 AM, Jacques Dainat <[hidden email]> wrote:

Dear all,

It is not the first time I see that, but I have an annotation launched with the option keeps_pred=0 that contains gene models with AED score equal to 1. As far as I understand, AED score to 1 means there is no evidence support. So it should be purely abinitio without any line of evidence in front of this prediction.
But as it is written in “Michael S-Cambell et al., Genome Annotation and Curation Using MAKER and MAKER-P,  Current Protocols in Bioinformatics, 2014” page 4.11.36 about the the keep_preds option:
"MAKER rejects models that do not have at least some form of evidence support.” Setting keep_preds to 1 “remove the evidence support requirement”.

So, how should I understand my results? Some predictions with low support (but support anyway) can have AED score equal to 1? Or some purely abinitio prediction without support at all can anyway be selected even when keep_preds=0 is set up? 

Best regards,

Jacques Dainat

_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org


_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
Reply | Threaded
Open this post in threaded view
|

Re: keep_preds=0 but some gene models have anyway an AED equal to 1.

Jacques Dainat-4
Hi, 

Thank you for your reply, nevertheless it doesn't it answer my question. Either I didn't express myself well enough or I don't get something obvious from your answer.

I will try to rephrase the problem.
My problem is that I am setting keep_phred=0 and that I obtain AED of 1. Can unsupported data can be selected with keep_phred=0? what I understood is that it is not the case so do you have any idea of why I have those AED equal to 1?

Thank you again for your help.
Best regards,

Jacques


On 4 May 2018, at 20:30, Carson Holt <[hidden email]> wrote:

By default MAKER will not let models through without at least some degree of evidence overlap (AED < 1). This is because ab initio predictors overcall (sometimes by as much as a factor of 10, i.e. 10 false positives for every true positive). You can dial in the minimum AED support using the AED_threshold. But if yo8u want to keep everything that does not have overlap with a better supported model then setting keep_pred=1 will allow even unsupported models to be maintained (1 for yes and 0 for no). Situations where you may want to do this include passing in old annotation datasets or working on organisms with very high genedensity and low ab initio false positive rates (many species of fungi meet this criteria).

—Carson


On May 2, 2018, at 5:55 AM, Jacques Dainat <[hidden email]> wrote:

Dear all,

It is not the first time I see that, but I have an annotation launched with the option keeps_pred=0 that contains gene models with AED score equal to 1. As far as I understand, AED score to 1 means there is no evidence support. So it should be purely abinitio without any line of evidence in front of this prediction.
But as it is written in “Michael S-Cambell et al., Genome Annotation and Curation Using MAKER and MAKER-P,  Current Protocols in Bioinformatics, 2014” page 4.11.36 about the the keep_preds option:
"MAKER rejects models that do not have at least some form of evidence support.” Setting keep_preds to 1 “remove the evidence support requirement”.

So, how should I understand my results? Some predictions with low support (but support anyway) can have AED score equal to 1? Or some purely abinitio prediction without support at all can anyway be selected even when keep_preds=0 is set up? 

Best regards,

Jacques Dainat

_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org



_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org
Reply | Threaded
Open this post in threaded view
|

Re: keep_preds=0 but some gene models have anyway an AED equal to 1.

Carson Holt-2
You should not see AED=1 in the annotations (unless you supplied features to model_gff those are always maintained). But you will see AED=1 in the evidence track. Make sure you are looking at features with a source tag of ‘maker’ and type of gene/mRNA/exon/CDS, and not type match/match_part. The match/match_part features are reference features in the evidence track. The reference features will also have their own fasta file. The only fasta you should use are the maker.proteins.fasta and maker.transcripts.fasta not the snap_masked.protein.fasta for example.

—Carson

On May 7, 2018, at 5:25 AM, Jacques Dainat <[hidden email]> wrote:

Hi, 

Thank you for your reply, nevertheless it doesn't it answer my question. Either I didn't express myself well enough or I don't get something obvious from your answer.

I will try to rephrase the problem.
My problem is that I am setting keep_phred=0 and that I obtain AED of 1. Can unsupported data can be selected with keep_phred=0? what I understood is that it is not the case so do you have any idea of why I have those AED equal to 1?

Thank you again for your help.
Best regards,

Jacques


On 4 May 2018, at 20:30, Carson Holt <[hidden email]> wrote:

By default MAKER will not let models through without at least some degree of evidence overlap (AED < 1). This is because ab initio predictors overcall (sometimes by as much as a factor of 10, i.e. 10 false positives for every true positive). You can dial in the minimum AED support using the AED_threshold. But if yo8u want to keep everything that does not have overlap with a better supported model then setting keep_pred=1 will allow even unsupported models to be maintained (1 for yes and 0 for no). Situations where you may want to do this include passing in old annotation datasets or working on organisms with very high genedensity and low ab initio false positive rates (many species of fungi meet this criteria).

—Carson


On May 2, 2018, at 5:55 AM, Jacques Dainat <[hidden email]> wrote:

Dear all,

It is not the first time I see that, but I have an annotation launched with the option keeps_pred=0 that contains gene models with AED score equal to 1. As far as I understand, AED score to 1 means there is no evidence support. So it should be purely abinitio without any line of evidence in front of this prediction.
But as it is written in “Michael S-Cambell et al., Genome Annotation and Curation Using MAKER and MAKER-P,  Current Protocols in Bioinformatics, 2014” page 4.11.36 about the the keep_preds option:
"MAKER rejects models that do not have at least some form of evidence support.” Setting keep_preds to 1 “remove the evidence support requirement”.

So, how should I understand my results? Some predictions with low support (but support anyway) can have AED score equal to 1? Or some purely abinitio prediction without support at all can anyway be selected even when keep_preds=0 is set up? 

Best regards,

Jacques Dainat

_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org




_______________________________________________
maker-devel mailing list
[hidden email]
http://box290.bluehost.com/mailman/listinfo/maker-devel_yandell-lab.org