ID attribute on CDS maketh error while running liftoff

[Yoonsung1203]

Hello,

while I was utilizing the recent GFF file for gene annotation on my genome assembly with Liftoff, the error raised as below:

UnboundLocalError: cannot access local variable 'feature_db' where it is not associated with a value

It seems the ID attribute on CDS features of chromosome Y (ex- ID=XP_011529739.1 on chromosome Y) is the cause of this error.
The multiple CDS lines on the GFF file exist with the same ID attribute, which causes non-unique IDs while loading the GFF file.

After removing the ID tags from the CDS feature, liftoff runs without error.

If it were not the case that you intended to mark something else, it would be better to fix those lines.

Thank you.

[diekhans]

A series of CDS records are allow to have the same id. This is a misfeature of GFF3 called "discontinuous features". I believe that this was put to support using a feature name of "CDS" for the CDS region of an exon when the sequence ontology feature is actually CDS_region. This causes a lot of pain, however chm13v2.0_RefSeq_Liftoff_v5.2.gff3.gz looks valid. I would suggest filing a bug against liftoff: https://github.com/agshumate/Liftoff You can most likely get liftoff to run by using gffread to convert the file to GTF