Confused between training input and prediction input

Discussions about WebAUGUSTUS, the Web Service of AUGUSTUS.

Moderator: bioinf

Post Reply
User avatar
katharina
Site Admin
Posts: 530
Joined: Wed Nov 18, 2015 6:14 pm
Location: Greifswald
Contact:

Confused between training input and prediction input

Post by katharina »

Originally posted by Chirag Parsania in the old forum on 13.08.2015 - 12:54

Hi,
I am using augustus web server to predict genes. I have trained augustus using verified ORFs also provided cDNA sequences while training. As an output of training I got optimised parameter for my gene structure. Along with that, I also got predicted genes (both ab initio and with hints) from supplied genome file with the newly optimized AUGUSTUS parameters. Till now everything is fine.

My question is as I already got predicted genes ab initio and with hints, do I need to run prediction as well. ?

For a try I also did prediction with the parameter which I got after training. Again I supplied same cDNA file which I supplied with training program. As an output I got gene structure but the number of genes predicted this time is not matching with number of genes I got after training algorithm.

Kindly help me to solve this confusion. I am using cufflink generated transcripts as cDNA file.
User avatar
katharina
Site Admin
Posts: 530
Joined: Wed Nov 18, 2015 6:14 pm
Location: Greifswald
Contact:

Re: Confused between training input and prediction input

Post by katharina »

In principle, if you don't want to add more extrinsic evidence, and if you don't want to alter prediction options, you don't need to run gene prediction, again.

It is possible that WebAUGUSTUS prediction will produce slightly different results. As it is possible that two seemingly identical augustus runs produce slightly different results.
Post Reply