You really can't do that with something like protein folding simulations (I would know, I've worked in the field for three years...).  In fact, most scientific problems involve calculations applied to stochastic problems where they don't necessarily know what the results will be.  The only way I can think of is to just use a ripple like system where whatever at home gives you credits based upon your solutions being provided to them.  But it's not necessarily a bad answer to this problem, it's better than folding and getting nothing but points on a leaderboard.

If opencoin just released their source code you could directly port it and then use it to make a network like ripple for your coins -- but they haven't.