Genome transcript annotation submissions are in GTF format (see GTF2.2 and Ensembl GFF/GTF File Format).
Model files must be gzip-compressed and named models.gtf.gz.
Only exon features are used, and exons of a given transcript are linked together by the transcript_id attribute. Gaps of at least XX bases between exons are assumed
to be introns. Smaller gaps are closed, combining the exon features. The evaluation ignores other features
that are provided. However, they will still be validated for syntax and consistency.
The standard GTF fields have the following restrictions:
exon is used. Other types of records are ignored.+, or - on exon features. A strand of . is not allowed. Other feature types may be validated.. on exon features, but not valid. Other feature types may be validated but are not required.transcript_id - Required for all exon features and assigned by the submitter.gene_id - Required for all exon features. While not evaluated by LRGASP, it is unclear if GTF requires this. Some tools, such as the UCSC Genome Browser and command-line tools, will fail without gene_id. The assigned gene_id does not have to represent an actual gene. However, it must not reflect impossible genes, such as those on different chromosomes. If gene assignment is not supported, it is suggested that each transcript is assigned its own gene_id, which may be the same as the transcript_id.reference_transcript_id - Optional, used to indicate the transcript is a reference call for the specified reference transcript.reference_gene_id - Optional, used to indicate the gene is a reference call for the specified reference gene. This attribute is used without a reference_transcript_id attribute if the models are not assigned to a specific transcript ofthe gene.All of the attributes specified above must be repeated on all exons of a transcript with the same value. An example models submission GTF file is at models.gtf.