iteratively running maker...

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

iteratively running maker...

Jamie
Dear MAKER community,

I am currently at the Arthropod genomics conference in Kansas City. I've just had a conversation with someone here that raised an important question about how to use MAKER and suggests I may be very confused about an important point -- particularly, when/why would I want to repeatedly run MAKER on the same data set.

I was under the distinct impression that, once all of the final parameters & data were in place there was only a need to run MAKER once to get a final result.  For instance, once ESTs, BLAST dbs, training models, etc were established, only a single run of MAKER is needed/recommended.  The only scenario I thought iterative MAKER runs were needed was when generating training data for the SNAP HMM, when it is suggested to predict with est2genome, train the HMM, then run MAKER with SNAP, and train again, etc. (as the tutorial describes).  

However, my colleague is convinced that iteratively running MAKER with all of the final params & data in place will improve the predictions.  For my part, having actively used MAKER recently, it is not clear to me how to do this (where would you provide the output from a previous run in the ctl file?) and I feel like I've never seen this iterative approach recommended anywhere.  

Can someone who knows please say whether my colleague is confused or if I'm missing important?  If it is me, more details on how/why to run MAKER iteratively to produce a final annotation would be much appreciated.

Thanks

Jamie
_______________________________________________
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: iteratively running maker...

Marvin B Moore
Jamie,

You are correct.  If you don't retrain any of the gene finders and don't tweak any other input parameters or files then there is no need to run MAKER more than once.  In fact, I'm pretty sure (and others can correct me if I'm wrong) that if you don't change any input parameters or files that you would be guaranteed to the get the exact same results.  I don't think there stochastic processes in the pipeline.

B


On Jun 10, 2011, at 3:38 AM, Jamie Walters wrote:

Dear MAKER community,

I am currently at the Arthropod genomics conference in Kansas City. I've just had a conversation with someone here that raised an important question about how to use MAKER and suggests I may be very confused about an important point -- particularly, when/why would I want to repeatedly run MAKER on the same data set.

I was under the distinct impression that, once all of the final parameters & data were in place there was only a need to run MAKER once to get a final result.  For instance, once ESTs, BLAST dbs, training models, etc were established, only a single run of MAKER is needed/recommended.  The only scenario I thought iterative MAKER runs were needed was when generating training data for the SNAP HMM, when it is suggested to predict with est2genome, train the HMM, then run MAKER with SNAP, and train again, etc. (as the tutorial describes).  

However, my colleague is convinced that iteratively running MAKER with all of the final params & data in place will improve the predictions.  For my part, having actively used MAKER recently, it is not clear to me how to do this (where would you provide the output from a previous run in the ctl file?) and I feel like I've never seen this iterative approach recommended anywhere.  

Can someone who knows please say whether my colleague is confused or if I'm missing important?  If it is me, more details on how/why to run MAKER iteratively to produce a final annotation would be much appreciated.

Thanks

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


Barry Moore
Research Scientist
Dept. of Human Genetics
University of Utah
Salt Lake City, UT 84112
--------------------------------------------
(801) 585-3543





_______________________________________________
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: iteratively running maker...

Carson Hinton Holt
Re: [maker-devel] iteratively running maker... That is correct.  Iteratively running without changing anything will not produce better results.  Iteratively running in the tutorial was used to train the gene finder using incomplete evidence from ESTs, and we were able to show that models improved on the second round after retraining SNAP.   It was an exercise in training gene finders for species using MAKER where you have no preexisting parameters or gene models.

--Carson


On 6/10/11 6:06 AM, "Barry Moore" <bmoore@...> wrote:

Jamie,

You are correct.  If you don't retrain any of the gene finders and don't tweak any other input parameters or files then there is no need to run MAKER more than once.  In fact, I'm pretty sure (and others can correct me if I'm wrong) that if you don't change any input parameters or files that you would be guaranteed to the get the exact same results.  I don't think there stochastic processes in the pipeline.

B


On Jun 10, 2011, at 3:38 AM, Jamie Walters wrote:

Dear MAKER community,

I am currently at the Arthropod genomics conference in Kansas City. I've just had a conversation with someone here that raised an important question about how to use MAKER and suggests I may be very confused about an important point -- particularly, when/why would I want to repeatedly run MAKER on the same data set.

I was under the distinct impression that, once all of the final parameters & data were in place there was only a need to run MAKER once to get a final result.  For instance, once ESTs, BLAST dbs, training models, etc were established, only a single run of MAKER is needed/recommended.  The only scenario I thought iterative MAKER runs were needed was when generating training data for the SNAP HMM, when it is suggested to predict with est2genome, train the HMM, then run MAKER with SNAP, and train again, etc. (as the tutorial describes).  

However, my colleague is convinced that iteratively running MAKER with all of the final params & data in place will improve the predictions.  For my part, having actively used MAKER recently, it is not clear to me how to do this (where would you provide the output from a previous run in the ctl file?) and I feel like I've never seen this iterative approach recommended anywhere.  

Can someone who knows please say whether my colleague is confused or if I'm missing important?  If it is me, more details on how/why to run MAKER iteratively to produce a final annotation would be much appreciated.

Thanks

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


Barry Moore
Research Scientist
Dept. of Human Genetics
University of Utah
Salt Lake City, UT 84112
--------------------------------------------
(801) 585-3543






Carson Holt
Graduate Student
Yandell Lab
http:/www.yandell-lab.org/
Eccles Institute of Human Genetics
University of Utah

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