Of course, with bitcoin it's not possible (unless you want to crack sha-256 hashed passwords maybe?), I was wondering if by designing the bitcoin algorithm differently that might have been possible.

For instance, to stay with the protein folding example, maybe bitcoin could have used an algorithm that used similar logic as the one used by the folding@home project but with different input data and expected results. Once the hardware is no longer needed for mining it can be easily repurposed for research.