[Bioperl-l] Re: [SO-devel] Re: GFF3 preliminary

[Chris Mungall]

I agree with RIchard that there should be a one-to-one relationship
between features and lines.

Also, we discovered at the meeting that we can't represent the full gene
model by attaching coding_start/coding_end directly to mRNA, as this makes
representation of dicistronics problematic.

What about this:
CDS/ORF as children of mRNA with n:n cardinality between CDS and mRNA. The
CDS feature indicates the coding_start, coding_end (actual coding exons
are implicit from exons + mRNA) so coding_start, coding_end become
redundant and are thus removed from the 'exchange-compliance' part of
SOFA.

For the richer part of SOFA, we allow coding_start, coding_end but there
are restricted to being 3bp long (representing Met and stop)