Dr. Daniel MacArthur returns from hiatus, and observes that deCODE is now in the advanced stages of corporate death. Pointers to a Newsweek article and other bloggers at Dr. MacArthur's post. Over at Gene Expression Classic p-ter points out that Small genetic effects do not preclude drug development (one of the rationales given for why deCODE couldn't translate associations in $$$).
Also, for historical perspective, check out this 2002 article about deCODE.
The assumption that even genes with common disease variants would have made good drug targets is bogus anyway, though great for tricking venture capitalists and funding agencies. The idea that to treat a disease you treat the genetic lesion is certain to fail as a biological strategy in all but a few special cases. Maybe it's worth finding the few, but gene effects are often developmental, and genes are almost always pleiotropic. The "disease" is not the mutation, it's downstream effects, often very downstream. There are so many quirky aspects of a given molecule (protein, lipid, nucleic acid, whatever) that might make it a good or bad drug target, I would guess that another component of the disease pathway or system is almost always going to be a better candidate that the one protein whose function is compromised or altered by a coding sequence difference. Find causal mutations can put you on track to understanding the disease at a systems level, but it's not a solution.
That said, I think gene therapy still holds a lot of promise for loss of function in constitutively required genes, like those involved in some neurodegenerative diseases.