[Bioperl-l] Re: Project

Jason Stajich jason@chg.mc.duke.edu
Wed, 11 Jul 2001 13:52:29 -0400 (EDT)


On Wed, 11 Jul 2001, Doug Hyatt wrote:

> Jason Stajich wrote:
> 
> > hmm, it looks like the GRAIL/GRAILEXP author Doug Hyatt was going to get
> > started on it, but I'm not sure where he is on it - if you want to contact
> > him directly he may or may not want help.
> >
> 
> Hi,
> 
> I've been on vacation and just got back.  I was still in the
> "figuring out bioperl" stages when I left.  My primary difficulty in figuring out
> 
> what to do is that Grail does exons, gene models, repeats, cpgs, polyas,
> promoters, frame shifts, etc.  so I wasn't sure if I was just going to do
> the grail2 exons or be really ambitious and do something for the
> whole package.
> 

We are in the midst of trying to build the Gene object in
Bio::SeqFeature::Gene, Hilmar Lapp had been the lead on this, I think we
were going to evaluate the next step in the process at BOSC next week. But
this is might be what we should ultimately be producing from the parsing
of the Grail reports.  I'm not 100% sure whether or not the
Prediction::Exon Prediction::Gene are rich enough or if they will be
merged into the SeqFeature::Gene objects as we get a better handle on the
data objects.  This is somthing we'll surely need input on from a variety
of sources.


> In addition, the example GrailEXP output file doesn't contain
> any alignment/evidence information (which again was something
> I had to figure out what to do with...)  Maybe I should commit
> some more example files for Grail1.3 and GrailEXP.
> 
yes! yes! for sure.  I just did a quick one-off to give us a starting
point.

> I guess the simplest thing to do with Grail would be to have
> the "next_prediction" routine return a Prediction::Exon object.
> I was looking at maybe implementing the "next_feature" command
> and getting lost in all the seqanalysisparser and other files :)
> Grail 1.3 does do gene modeling (if you submit exon clusters and
> specify a range to assemble across), but by far its most common
> usage is just to get the Grail2 exons.
> 
> Simplest thing for GrailEXP would be to have "next_prediction"
> return a Prediction::Gene object.  I still have no clue what to do
> with the evidence/alignments that come out of the program yet.
> Still learning....
>

We return exons from the MZEF module because that is all that MZEF
predicts, we should be returning genes from GRAIL since GRAIL tries to
predict genes.

I'm not sure I have the complete answer here, but the next_prediction
should return something that is Bio::SeqFeatureI - if we can't wrap all
the necessary data in the Prediction::Gene object we need to evaluate
their structure and decide how to create the proper object here.
 
> If someone else wants to write these who has a better understanding of
> what bioperl objects to use, I'm happy to just assist/answer
> questions... Otherwise, I'll plod along on it myself (but I can't
> promise speed--I'm still a bioperl newbie).
>

> regards,
> doug
> 

Jason Stajich
jason@chg.mc.duke.edu
Center for Human Genetics
Duke University Medical Center 
http://www.chg.duke.edu/