Repository navigation
Only noncoding annotations returned despite protein-coding genes in reference DB #6
Description
Hi, thanks for this tool!
I’m trying to use this tool, but I'm running into an issue where all annotations are classified as noncoding, even though my reference database clearly contains protein-coding genes.
Interestingly, this seems related to the already reported (but unresolved) issue:
#3
Instead of expected annotations like:
- CDS
- exon
- UTR
I only get:
- distnoncoding_intron500
- intergenic
- noncoding_exon
- proxnoncoding_intron500
I generated my own gffutils database from GENCODE:
- Source: GENCODE Comprehensive gene annotation
- Release: 49 (GRCh38.p14)
db = gffutils.create_db(
annotation_file, dbfn=db_file, force=force,
keep_order=True, merge_strategy='merge', sort_attribute_values=True,
disable_infer_genes=False,
disable_infer_transcripts=False
)I verified that the database contains protein-coding genes: and got ~20,000 genes across ~290,000 transcripts.
Question:
Could this be related to how gene/transcript biotypes are parsed or expected by the annotator?
For example:
- Does the tool require specific attribute keys (e.g.
gene_biotypevsgene_type)? - Is there a compatibility issue with newer GENCODE releases?
Thanks a lot for your help!
Activity
byee4 commentedon Mar 27, 2026
Hi, it's possible it could be related to the open issue, but unless Gencode has recently changed their format, the annotations should have those gene types. Could you help me with a few lines of your unannotated file, so I can check? Is your data stranded?
schilau commentedon Mar 30, 2026
Of course! I’m not able to upload a .bed file directly, so I simply changed the file extension.
Yes, my data is stranded!
Do you need any additional information? Thanks a lot!
test.bed.txt
byee4 commentedon Oct 5, 2026
I tested the attached 168-row BED against a gffutils DB made from the GENCODE v49 comprehensive GTF. I retained every feature for the genes overlapping the supplied chr12/chr14 intervals and used the DB creation options from the issue.
With
--format gencode(the CLI default), the currentmastercode assigns every row to a protein-coding gene: MSRB3 (4), HMGA2 (126), or DDHD1 (38). The selected regions are 1123utr, 26CDS, 105utr, and 20 intronic calls.Using
--format ensemblwith that same GENCODE database reproduces the noncoding-only pattern from the report: 150noncoding_exon, 15distnoncoding_intron500, and 3proxnoncoding_intron500. GENCODE v49 usesgene_type/transcript_typeattributes and a genericUTRfeature, so--format gencodeis the appropriate setting. PR #10 improves the biotype-key fallback if a different format is selected, but the correct format is still needed for 5′/3′ UTR classification.Could you share the exact
annotatorcommand you ran (especially--format), and confirm that the database was created fromgencode.v49.annotation.gtf.gz? I’m leaving this open pending that detail.