Why are gene models from JGI non-sense when using AGAT to to extract proteins?

Question

[O.rka]

https://genome.jgi.doe.gov/portal/pages/dynamicOrganismDownload.jsf?organism=Phypa1_1

I have 3 files:

• Assembly: Physcomitrella_patens.1_1.allmasked.gz
• GFF: Phypa1_1.FilteredModels.gff.gz
• Proteins: proteins.Phypa1_1.FilteredModels.fasta.gz

There are no CDS sequences so I'm going to try and recreate the protein file and also the CDS file. However, when I use agat to extract the sequences, I get nonsense:

seqkit seq Physcomitrella_patens.1_1.allmasked.gz > Physcomitrella_patens.1_1.allmasked.fasta

agat_sp_extract_sequences.pl -f Physcomitrella_patens.1_1.allmasked.fasta -g Phypa1_1.FilteredModels.gff.gz -c agat_config.yaml -p -o test_proteins.fasta

My proteins are full to stop codons:

>agat-rna-1 gene=agat-gene-15249 seq_id=scaffold_1 type=cds
*SSKLQKHRAQVEHSVAHVHA*M*RWDFFGRGLGYGEEARRRNSTSFISHCRRDALQNFY
HSFTFEMRTKSVPTGDHTPEGT*DSYPLRLMGAKKAW*SVSKFTVVSHLR*QMQTLNAGT
QSRVIVGRRNNSVTELLHGMEIN*WFSTQSVRSLVAAKLQISSGGRKIRLR*ASGLILFP
LIG*Q*T*EIRLLFLPVINPLQALRPSPKSTLGKDMIHCEGQ*LLQADVMI*KRNYISMM
QVQ*LEYHCGSFRAIELLLATHLIP*WQPVQLIGA

The actual sequence for this record should be the following:

>jgi|Phypa1_1|63627|fgenesh1_pg.scaffold_1000001
MKFKAAKAQSPSGTFCGSCACMNVKMGFFWTGVGLWGRSKEEKQHKLHKSLSKRCIAEFL
PQFHIRDADEVRSNRRPYTGGDVRLLPTEVNGGEEGLVICLQVHRSLPSSVADADSECWD
AIPRYRWKKEQLSHRVVARDGNQLMVFDAICEVTGCCKAANIFRRSEDSVKVSFRLDFIS
AYRVTMNLRNSIVVPPCDQSTASITTLSEIHSRQRHDPLRGSVIAASRCHDLKKELHFHD
ASPITRVPLWQLSGHRVASCNPSHSLMAASSVNWS*

The genetic code is trans_table 1

[Juke34]

Are you sure the fasta file is sync with the annotation file? I.e the annotation gff/gtf has been done using this fasta?

[O.rka]

It's difficult to be 100% sure but there's only one genome assembly. I've updated the question with the screenshot.