As I pointed out to shmerl above, sequence space is very large, and in general sparsely populated. A functional gene product and regulatory sequence is often thousands of base pairs. 4^1000 = 10^600. There are only 10^80 molecules in the universe. So if you have done truly de novo design, you can be pretty sure it doesn't exist somewhere else by random chance.
That being said, I think that even if you take a natural gene and modify it meaningfully - that should be patent-able. For instance, if you take an enzyme and engineer a version that doesn't degrade as readily, or has much better activity, that should be patent-able, don't you think?
EDIT: There's a good deal of prior art regarding numbers. As for patenting molecules, we allow that all the time... drugs, compounds, etc. Because they are discovered as I describe above.
Again, getting into arguments of making molecules patentable because there are so many combinations available is bizarre. Why not make numbers patentable as well - there is an infinite amount of them.
That being said, I think that even if you take a natural gene and modify it meaningfully - that should be patent-able. For instance, if you take an enzyme and engineer a version that doesn't degrade as readily, or has much better activity, that should be patent-able, don't you think?
EDIT: There's a good deal of prior art regarding numbers. As for patenting molecules, we allow that all the time... drugs, compounds, etc. Because they are discovered as I describe above.