And it's an interesting decision. As mentioned here, the company's last stand was on questions of patentability, specifically the written description requirement. Well, the appeals court has ruled Monday morning, and Ariad's '516 patent does, finally, appear to be invalidated. There's more at Patently Obvious, who seem to be among the first with this story.
From what I can see, the court's decision makes it clear that there really has to be a description sufficient for one skilled in the art to reproduce an invention, and that stating your hypothesis isn't enough to meet this requirement. So in Ariad's case, claiming all sorts of (not yet existing) things to modulate NF-kB function doesn't fly, because they don't actually tell anyone how to do that, just how they wanted to own it if and when someone does. The written description requirement, the court holds, doesn't mean that you have to actually reduce something to practice (although I'd have to say, from my own perspective, that it most certainly would be a good idea to do so if you can), but you have to show how that could be done. "Patents are not awarded for academic theories, no matter how groundbreaking or necessary to the later patentable inventions of others" is a key quote.