I don't really have any great feeling either way about in which coordinate set
the CDS description should be (although working in coordinates of what is
essentially a subsequence of a subsequence, while more correctly representing
the biology, is harder to get your head around).
What I would like is to see this all properly implemented so that we can get
around the awful fudge that is prevalent now in the sanger human acedbs whereby
to describe the translated portion of a transcript two subsequence objects are
made, the transcript itself and then the translated (CDS) object. To make this
look nice the views of the two objects are tweaked so that they are in different
colours and the translation lies on top.
What seems nicer to me would be to only make transcript subsequences, and to
describe the translation of these using CDS Int Int in whatever coordinates are
deemed appropriate, and to add in a function to colour or fill in the translated
portion of the transcript in the fmap display. (BTW I presume that the system
allows description of the translation of polycistronic messages?)
I believe that this must be what was originally intended but it isn't getting
used like this in the human databases at least.