Stop codon included in CDS features of ORFanage GTF output

[nuoxuxu]

Hi Ales,

Thank you so much for developing ORFanage! I have some suggestions for the output format. Currently the ORFanage output GTF file includes stop codons in the CDS features. This could potentially cause problems in downstream tools since it is assumed that the GTF file format does not include stop codons in the CDS feature (see https://agat.readthedocs.io/en/latest/gxf.html#main-points-and-differences-between-gtf-formats).

To adjust the CDS coordinate, I used agat_sp_add_start_and_stop to add the "stop_codon" feature, and then used agat_convert_sp_gff2gtf to convert the resulting gff3 file to a gtf file, from which the CDS coordinates are adjusted so that stop codons are not included.

It would be very helpful to state somewhere in the documentation that the ORFanage output gtf includes stop codons in the CDS features.

Another suggestion is that the current ORFanage gtf output does not included "gene_id" for features like "exon" and "CDS", this could potentially cause problems in downstream tools such as agat, since it is assumed that all features of a GTF file have a gene_id and a transcript_id attribute (see here).

Best,
Nuo

[alevar]

Hi Nuo,

Thank you for the thoughtful suggestions! In hindsight, I agree that stop codons should not be included in the output (or at least there should exist a CLI flag to include/exclude stop codon positions in the CDS features). The choice to keep them by default was motivated by added ability to quickly scan for incomplete proteins in GTF files - if stop codons are included in CDS, simple search for "*" or "." in translation via gffread would reveal incomplete sequences. I believe GENCODE opts for including stop codons in the output as well in contrast to NCBI - it is very unfortunate how loose the GFF/GTF specifications are and how difficult that makes enforcement of a single standard. I will work to refine handling of stop codons in the CDS records for ORFanage and will keep you posted in this thread on the changes when they are available.

Regarding the gene_id - I agree, it might be useful to have the gene_id propagated down from transcript level to the exon and CDS levels where possible and should not present conflicts with downstream methods. I will make this change shortly as well.

Again, thank you for the suggestions and I will update this thread when the changes are ready to be used!

Ales