Dear Tim
you could add an exon class to the current acedb models
and XREF those exons to the genes
but this is going to be quite heavy becuase the human genome has
at least 1M exons
this is why i was proposing to add the exon relational identifier
in the constructed type
This is ligth for the system and may be sufficient
disk storage space for that extra value is non significant
and data concistency is anyway an external problem