[maker-devel] MAKER annotations

[Carson Holt]

If your organism is biased toward short genes with few introns, you might
also want to try something like GeneMark.  Not all gene predictors work well
on all kinds of gene structures, so I often will run multiple and then drop
those that perform poorly.

Thanks,
Carson


From:  Sivaranjani Namasivayam <ranjani at uga.edu>
Date:  Thursday, 20 December, 2012 1:33 PM
To:  Carson Holt <carsonhh at gmail.com>
Subject:  RE: [maker-devel] MAKER annotations

I am annotating a Apicomplexan organism.  When I retrained I used both SNAP
and Augustus. I will lower the AED score to 0.5 and check if that improves
the results.
Thanks for suggestion!

Ranjani

From: Carson Holt [carsonhh at gmail.com]
Sent: Thursday, December 20, 2012 1:16 PM
To: Sivaranjani Namasivayam; maker-devel at yandell-lab.org
Subject: Re: [maker-devel] MAKER annotations

Don't use an AED filter of 0.  It's too strict and will introduces certain
artifacts like partial genes.  Lower than 0.5 is probably sufficient for
training.  Also if in your first round you use Augustus, and the second
round you used SNAP instead, you may get poorer annotations because Augustus
is performing better than SNAP.  SNAP might need more training as well,
especially if the low AED filter introduced partial gene artifacts.    What
kind of organism are you annotating?

--Carson


From: Sivaranjani Namasivayam <ranjani at uga.edu>
Date: Thursday, 20 December, 2012 12:21 PM
To: "maker-devel at yandell-lab.org" <maker-devel at yandell-lab.org>
Subject: [maker-devel] MAKER annotations

Hi,

I have a query regarding MAKER predictions. I made a a first round of
predictions with RNAseq data from 454 and Illumina as EST evidence and
proteins sequences from related organisms as proteins evidence. I also used
augustus trained on a closely related organism as the ab-initio gene
predictor.
I got ~4000 maker annotations (I expect around ~10,000 genes)

I used the annotations with AED score of 0 to train SNAP and used the HMM
when retraining MAKER. This time I got ~8,000 genes but when I examined some
annotations they appeared to be have worsened when compared to the initial
predictions: that is, genes were getting split, UTRs were not annotated and
the original predictions were modified but not for better.

Would you have any suggestions on how I can improve the annotations.

Also, is there a way to tell MAKER to produce gene models for all the
cufflinks assembled transcripts(that is, for the data set I provide as EST
evidence).

Thanks,
Ranjani
_______________________________________________ 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/20121220/0d7093ff/attachment-0003.html>