Dear community,

I annotated a fungal genome using BRAKER3 (--gff3 --fungus). The resulting GFF3 is valid (passes gt gff3validator) and contains mRNA features with IDs like g6.t1, all with product=hypothetical protein (expected at this stage).

I then ran DIAMOND blastp of the predicted proteins against a custom UniProt/TrEMBL-based FASTA database. The DIAMOND output is standard outfmt 6, with qseqid matching exactly the mRNA IDs in the GFF3 (e.g. g6.t1), and multiple queries clearly have strong hits with informative descriptions (e.g. Subtilase).

I tried to add functional annotations using

agat_sp_manage_functional_annotation.pl -f braker.gff3 -b diamond_blastp.out --db db.fasta

However:

AGAT never updates product=, which remains hypothetical protein for all mRNAs

No Note or Dbxref attributes are added either

error.txt reports many messages like:

“ERROR tr|XXXX|YYYY not found among the db! You probably didn't give to me the same fasta file than the one used for the blast”

This happens even though the subject IDs reported by DIAMOND do exist in the FASTA DB (confirmed by grep). I also tried: -restricting BLAST to best-hit-per-query -simplifying FASTA headers -ensuring the same FASTA was used to build the DIAMOND DB and passed to AGAT

The problem persists: AGAT parses the GFF but seems unable to reliably retrieve subject records from the FASTA DB, so functional annotation is not propagated.

Question: Is this a known limitation/bug of AGAT with UniProt-style headers (tr|…|…) or large FASTA databases? Is there a recommended way to add my DIAMOND annotations to BRAKER GFF3 (product field) when AGAT fails at this step?

Any advise or comment is appreciated. Thank you so much for your time!