[maker-devel] Fused gene problem, improvement in the Maker 2.27?

[Carson Holt]

Yes.  Barry gave a good overview.  The correct_est_fusion option basically
clips UTR when there are two neighboring genes that only overlap in the UTR
(so you still get both gene models).  Since the primary effect of falsely
merged mRNA-seq is overly long UTR this tends to fix many cases.  Of course
avoiding merging the mRNA-seq reads in the first place also works.  So using
Trinity's extra options to control that together with the correct_est_option
option in MAKER is probably the way to go.

I think you can lower pred_flank to 100, but below that you might start to
get weird behavior from the gene predictors (they need some upstream and
downstream sequence or the HMMs don't work well).

Thanks,
Carson


From:  Barry Moore <barry.moore at genetics.utah.edu>
Date:  Tuesday, 21 May, 2013 7:54 PM
To:  <Sean.Li at csiro.au>
Cc:  <maker-devel at yandell-lab.org>
Subject:  Re: [maker-devel] Fused gene problem, improvement in the Maker
2.27?

Hi Sean,

I think you want to be careful with dropping the pred_flank parameter too
low.  This controls how much flanking sequence (for a given cluster of
evidence) MAKER will pass to the gene predictor.  Some (maybe all?) of the
gene predictors have an initial state in their HMM for intergenic sequence
and if you do not have some intergenic sequence for them to consider first
they can't transition to their next state.  The correct_est_fusion option
can help (at the cost of losing some UTR annotations) - Carson will likely
give you a better description of the intricacies of the correct_est_fusion.

Don't know how you are assembling your RNASeq, but there is an option in
Trinity - I forget the name - that will instruct Trinity to be more
restrictive in merging neighboring clusters of reads into a longer
transcript and this can help as well.

B

On May 21, 2013, at 1:36 AM, <Sean.Li at csiro.au>
 wrote:

> Hi Carson,
>  
> We are currently working on the annotation of Helicoverpa genome project.
> Maker has been chosen as the preliminary tool for the task. By checking the
> annotation results by using maker 2.10, we saw some loci have the fusion
> problem: two separate neighbour genes are likely to be fused together and
> regarded as a single candidate output by maker. If we go further by looking at
> the outputs from each individual de novo algorithm, e.g. augustus or snap, the
> prediction was correct.  We are also using RNA-Seq assembly from cufflinks and
> some protein evidence data from closely related insects.
>  
> We noticed that the parameters ³pred_flank² in maker v2.10 and
> ³correct_est_fusion² in maker v2.27 might be useful for maker to decide when
> to merge models or not.  If possible, can you please explain what these two
> parameters can do with the predicted genes, RNA-Seq and protein evidence?
>  
> Also, our current plan is  to install maker 2.27,  train the algorithms to
> predict UTRs, enlarge the protein evidence datasets and input our previous
> annotations as model_gff. We are facing with an critical question: in which
> way we could effectively improve the gene fusing problem? 1) setting the
> pred_flank lower than 100? 2) turn the correct_est_fusion on? 3) anything
> else?  
>  
> Thank you.
>  
> With best regards,
> Xi (Sean) Li, Ph. D.
> 
> Bioinformatics Analyst, Bioinformatics Core,
> CSIRO Mathematics, Informatics and Statistics
> Phone: +61 2 6216 7138 <tel:%2B61%202%206216%207138>
> Address: GPO Box 664, Canberra, ACT 2601
>  
>  
>  
>  
>  
> _______________________________________________
> maker-devel mailing list
> maker-devel at box290.bluehost.com
> 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
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/20130521/92990665/attachment-0003.html>