[maker-devel] maker-devel post from david.carlson at stonybrook.edu requires approval

[Carson Holt]

Try both ways (run in same maker directory to avoid reprocessing of evidence alignments). Then look at the output in a browser. Really the best way to decide which works best id manual evaluation.

—Carson



> From: David Carlson <david.carlson at stonybrook.edu>
> Subject: Training Augustus using Transdecoder output
> Date: January 7, 2020 at 1:47:23 PM MST
> To: maker-devel at yandell-lab.org
> 
> 
> Hi Maker users and developers,
> I'm currently in the process of annotating a new genome, and I have a question regarding training Augustus.
> 
> I have RNA-Seq data from my organism, which I've mapped to my genome assembly with Hisat2 and assembled into transcripts with StringTie.  
> 
> I've used these transcripts to predict proteining coding genes with Transdecoder (using homology evidence from the swissprot and Pfam databases).
> 
> I was thinking about taking the Transdecoder output, (after removing sequences that are not full length) and using these as the initial training set for Augustus.
> 
> In the past, I've used Busco for the initial Augustus training, but my thought is that these Transdecoder predicted genes might be a better source of evidence for my particular species.
> 
> Do any of you have experience trying this approach, or thoughts regarding its merits?
> Thanks!
> Dave
> 
> -- 
> Dave Carlson
> PhD Candidate
> Ecology and Evolution Department
> Stony Brook University

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://yandell-lab.org/pipermail/maker-devel_yandell-lab.org/attachments/20200122/81763135/attachment-0003.html>