Retraining Augustus for UTR

Question

[Ric]

Hi, I used to retrain Augustus with help of BUSCO in the following way:

1. run_BUSCO.py -i NbV1Ch07.fasta -o augustus_busco -l /assembly/busco-embryophyta_odb9/ -m geno --cpu 1 --long -sp coyote_tobacco --augustus_parameters='--progress=true'
2. copied the files from /augustus/run_augustus_busco/augustus_output/retraining_parameters to /augustus-install-directory/config/species/BUSCO_augustus_busco_3750827137/
3. augustus --UTR=on --gff3=on --species=BUSCO_augustus_busco_3750827137 NbV1Ch07.fasta > output.gff

Unfortunately, I received the below error message:

The file with UTR parameters for BUSCO_augustus_busco_3750827137 does not seem to exist. This likely means that the UTR model has not been trained yet for BUSCO_augustus_busco_3750827137.
augustus: ERROR
        UtrModel::readProbabilities: Couldn't open file /augustus-install-directory/config/species/BUSCO_augustus_busco_3750827137/BUSCO_augustus_busco_3750827137_utr_probs.pbl

How could I add UTR training?

Thank you in advance,

[lieven.sterck]

I would suggest to deactivate the UTR-prediction all together ( --UTR=off ?) . UTR prediction is notoriously difficult (read: nearly impossible). The only way to get accurate UTR assignment is through transcript mapping.

Since BUSCO is using protein info I can't really see how it could provide accurate info for UTR prediction anyway.