I have thought long about biological computing. DNA computing seems too slow for the Terra-Flop crowd, but this one descibes DNA which uses computing sections to treat diseases in an algorithmic way. Very cool!

sacbee.com -- AP State Wire News -- Researchers find logic machines in biology

What I would like to see is a molecular level monostable vibrator circuit (flip-flop) which could be the basis for a molar computer.
I would think that a small molecule in a crystal array with 1) two stable states, 2) a way to fip the states and 3) a quick way to query the states would be a very fast way to store data a million times more dense than any magentic or solid state array.

I would guess that it would be like the old magentic core. A two signals could be used to query a 2D location. The signals themselves would generate a change in state all along the path, where two signales met the state would not change and a slightly higher impedance would be measurable. The system would have to reset itslelf back to original state by sending the two signals separately and the target spot would be in an inverse state. (This is way too hard to explain without diagrams.)
Anyway, I am waiting to see if womeone is actually doing this.

The only drawback is that the system will generate heat so that whatever molecule is used will have to be resistant to going pfut when the current goes up.