For a decade, Pfizer would not even try to develop any drug that did not meet RO5. It has to be said that the attrition-rate among candidates IS substantially lower. As sekio points out, their are other 'rules' that provide a 'druglikeness' value. One of the more obscure ones involves how achiral a compound is. Well, the way it was put to me was that a telephone directory might be 2000 pages but they use minimal space because the pages tessellate. Make 2000 paper hats and they will occupy significantly more space..
A drug that occupies more space but doesn't have to many rotatable bonds is more 'druglike'.
I've seen Javascripts that take a SMILES string and provide a 'druglikeness' value. Not something I would trust, but if something failed, I would at least look again.
Don't forget that the number of possible compounds with a MW under 300 is simply massive. We have only just grazed the surface. I have a friend who is a virologist and I just cannot work unless I can see the molecule in my head. Lucky I can manage the RO5!