Page 1 of 1
Creating training set from assembled RNAseq transcripts
Posted: Thu Nov 19, 2015 3:43 pm
by katharina
Originally posted in the old forum by Sam on 13.03.2013 - 13:12
Hi,
I have generated extrinsic hits from RNASeq data for my species. I want to test the enhanced prediction accuracy. Is it okay to test the improved prediction accuracy with the test set created from protein alignment with Scipio? If not then how can I create training set from assembled RNAseq transcripts?
Thanks!
Re: Creating training set from assembled RNAseq transcripts
Posted: Thu Nov 19, 2015 3:43 pm
by katharina
by katharina on 13.03.2013 - 13:45
Hi Sam,
you could create training gene structures from assembled RNA-Seq data using for example PASA (e.g. using
http://bioinf.uni-greifswald.de/webaugustus). The resulting quality highly depends on the quality of your assembly. I would probably stick with the Scipio-genes, but keep in mind that those genes must not necessarily be covered by RNA-Seq data.
Katharina
Re: Creating training set from assembled RNAseq transcripts
Posted: Thu Nov 19, 2015 3:43 pm
by katharina
by Sam on 13.03.2013 - 15:11
Thanks Katharina,
I will also stick with Scipio-genes. Is the hints generated from RNASeq data can also be used to re-optimize the Augustus parameters for trained species?
Sam
Re: Creating training set from assembled RNAseq transcripts
Posted: Thu Nov 19, 2015 3:43 pm
by katharina
by katharina on 13.03.2013 - 15:27
No, you cannot use hints to re-optimize AUGUSTUS parameters for protein coding genes.
Re: Creating training set from assembled RNAseq transcripts
Posted: Thu Nov 19, 2015 3:43 pm
by katharina
by katharina on 20.01.2015 - 15:33
We now provide and recommend BRAKER1 for training AUGUSTUS from RNA-Seq:
http://bioinf.uni-greifswald.de/bioinf/ ... index.html