In a message dated 6/16/2003 3:18:05 PM Eastern Daylight Time, writes:

my opinion is there should be an independent module that builds bioperl
sub_SeqFeature trees based on /gene, /locus_tag and the locations of the
splice sites in the mRNA/CDS features.


       Is splice site agreement between an mRNA and (nuc) CDS a sufficient
condition to infer a mRNA->prot link? It seems like necessary condition, and if you add the initial codon to the mRNA it seems sufficient. But does the fact that a codon is the first in some CDS guarantee that it is the first codon in that particular mRNA?
Eg maybe there are two mRNAs that initiate translation at two different
points, but both share the bits described in the CDS. Not sure how often
that would arise in practice.

Are the tags used in a consistent enough way across submissions to be of any use?

Cheers, -Stan