How to integrate Infernal output with tRNAscan SE 2.0

[Abieskawa]

Hi,
I am doing eukaryotic genome annotation, and I have done protein structural and functional annotation with Braker3 and GO and KO (it will create other txt file so gff file was not touched). I saw your team recommend to use infernal as generic prediction, and for specific type like tRNA, it is better to use other tools tRNAscan-SE. My question are:

1. How can I combine these two outputs (Infernal and tRNAscan-SE result)? I saw tRNAscan SE can be output as gff, which I think that can be easily integrate into braker3 gff result.
2. Can I ask Infernal not to predict tRNA and use tRNAscan SE instead?
3. Are there recommended tools that can integrate infernal/tRNAscan SE with protein predicted gene gtf/gff (in my case is braker3 result)

Sorry if this is stupid question.
Thanks~

[nawrockie]

I agree that tRNAscan-SE is a better tool for tRNA annotation than Infernal.
This page has instructions on how to annotate genomes with Infernal and Rfam:
https://docs.rfam.org/en/latest/genome-annotation.html

If you want to ignore Infernal/Rfam predictions for tRNAs, the easiest way is probably to just remove the lines with 'RF00005' or 'RF01852' (these are the Rfam accessions for the tRNA and tRNA-Sec models) from the .tblout files output from cmscan with a command like:
cat my.tblout | grep -v RF00005 | grep -v RF01852 > new.tblout

There's also a video tutorial that may be useful:
https://www.youtube.com/watch?v=NU63fazDZHs

You can convert infernal .tblout files to gff with this script:
https://github.com/nawrockie/jiffy-infernal-hmmer-scripts/blob/master/infernal-tblout2gff.pl

[Abieskawa]

Thanks for your kindly reply. But I have two more questions.
I had transformed the tblout this morning, and I found that the gff output might require modification.
I transformed the file with the following command line:
infernal-tblout2gff.pl --cmscan --fmt2 --source infernal beltfish_ncRNA_wo_tRNA.tblout > beltfish_ncRNA_wo_tRNA.gff
The below is the the output

I suppose the output should have ID=blablabla in the last column. and also a label tag of gff3 version. Like this from tRNAscan-SE output

And I wonder if the Parent tag is required or not in other RNA entries of infernal output gff... I am not familiar with the RNA field

Additionally, the classification of Rfam cannot match gff3 type format, for example, U1 described in first picture, is actually a kind of snRNA, and when I combine result from braker3, tRNAscan-SE, and infernal transformed output, AGAT warns me about that and remove the entries.

when I look the format from gff3 file of human reference

NC_000001.11	cmsearch	gene	6752723	6752821	.	+	.	ID=gene-LOC124904847;Dbxref=GeneID:124904847;Name=LOC124904847;description=small nucleolar RNA U13;gbkey=Gene;gene=LOC124904847;gene_biotype=snoRNA
NC_000001.11	cmsearch	snoRNA	6752723	6752821	.	+	.	ID=rna-XR_007067439.1;Parent=gene-LOC124904847;Dbxref=GeneID:124904847,RFAM:RF01210,GenBank:XR_007067439.1;Name=XR_007067439.1;gbkey=ncRNA;gene=LOC124904847;inference=COORDINATES: profile:INFERNAL:1.1.1;product=small nucleolar RNA U13;transcript_id=XR_007067439.1
NC_000001.11	cmsearch	exon	6752723	6752821	.	+	.	ID=exon-XR_007067439.1-1;Parent=rna-XR_007067439.1;Dbxref=GeneID:124904847,RFAM:RF01210,GenBank:XR_007067439.1;gbkey=ncRNA;gene=LOC124904847;inference=COORDINATES: profile:INFERNAL:1.1.1;product=small nucleolar RNA U13;transcript_id=XR_007067439.1

It has created hierarchy of exon, which seems not to be the original output.