dictyBase GFF3 has developed bunch of rough edges over the years and so do not plays well with third party tools. Here are the issues that we are aware of…
- Genes with multiple gene models could easily be confused as splice isoforms. It is particularly hard to separate in case of known isoforms. There is also no easy way to identify the primary gene models.
Pseudogenes are not represented with proper SO terms. The pseudogenes used gene SO term whereas the transcript is represented by pseudogene term.
More than handful of assembly gap coordinates falls outside the range of reference feature.
CDS features do not have a value for phase column, which is required in GFF3 specification.
Missing target attribute of EST features. This is becasue the aligned coordinates of EST sequences are not present in our chado database, probably left out during the loading process.
The gene models will be split into individual GFF3 file for canonical, non-coding, curated and predicted gene models. The reason for separation is that each gene model has its own customized data model particularly the way transcripts are represented. As a result, they have to be treated and exported separately.
The canonical will be the core GFF3 with all the chromosomes, contigs and genomic sequences. The rest of them (curated and predicted) will be supplemental and will contain only gene structures, no sequences and assembly features though. By default, all chromosomal features will be kept in a single file.
The pseudogene gene model will be fixed, the transcript will be psuedogenic_transcript whereas the exons will be psuedogenic_exon.
No more export of gap features at this point.
CDS features will not be exported at this point, would be included at a later point along with the phase column.
Rerun the EST pipeline and load alignments with EST sequence locations. It will ensure the presence of Target attribute. It will be available as separate alignment GFF3 file. However, until the rerun is finished the existing alignment GFF3 will be without the Target attribute.