OK. I know this is not a real answer to your question but I think you will not want to miss today's news, mentioned in the tweet by @Ananyo below. Ananyo Bhattacharya is Natures chief online editor. Of course this work did not lead to a new drug yet. But it will shake up drug research, and we now at least have the structures to do the prediction with:

RT @Ananyo: Biologists-don't miss @alohalizzie's gr8 news story on the new GPCR structure: Nobel prize-winning stuff http://bit.ly/oDvpqF

For instance for the reason @Ananyo gave in his second tweet about the topic:

RT @Ananyo: This is why I <3 crystallography. The GPCR structure cld explain how cholera toxin kills http://bit.ly/oDvpqF

Since you asked for the actual papers. Here it is: Rasmussen, S. G. F. et al. Nature http://dx.doi.org/10.1038/nature10361 (2011).

I think there is also another answer. If you really do find a drug using modelling (and from what I understand a number GPCR agonists and antagonists have been found in pharmaceutical industry in this way, using incomplete models for GPCR) you will not publish that in a scientific journal. You will instead verify your results and then get a patent. Publishing it in another way would be about the stupidest thing you could do. The publication would effectively ruin your chance to get a patent, and without the patent you would not be able to invest in the necessary toxicity studies and clinical trials. The latter is true since any competitor would be able to sell the drug at a lower price since he would not have to invest in the studies needed to get the new drug registered.

Since patents are in fact accessible you might want to search there.

There's a whole issue of Briefings in Bioinformatics dedicated to computational approaches to drug repositioning: http://bib.oxfordjournals.org/content/12/4.toc

Not drug discovery per se, but certainly an area that has shown a lot of promise in response to purely computational approaches.

I think Chris has a good point - protein structures are often what you need to have to even start guessing at drug structures, so its a lot easier to think of papers that are really the report of a protein structure.

A couple of examples:

HIV Protease structure in complex with peptide inhibitors. It was such an important target that they synthesized the protein on solid phase for the structure.

COX1 and 2 - primary targets for inflammation (I understand that the organic chemists built the COX inhibitors before the structure was done tho).

Influenza virus haemagglutinin - i think this actually helped spur the development of some flu drugs.

I would be interested to know if there were any purely insilico designed drug targets. Usually even insilico designed libraries of drug candidates are huge (> 100 million) and the computers use the structures of known drugs or of the protein target to throw out the first 80-90% then you have to just test the rest.

Identification of Selective Inhibitors of Cancer Stem Cells

http://www.sciencedirect.com/science/article/pii/S0092867409007818

Identified drug: salinomycin

One compound, salinomycin, reduces the proportion of CSCs by >100-fold relative to paclitaxel, a commonly used breast cancer chemotherapeutic drug.

My colleague suggested the following article for this answer:

Lessons learned from the development of an Abl tyrosine kinase inhibitor for chronic myelogenous leukemia http://www.jci.org/articles/view/9083

But they also used random screening to get the initial lead compound.

As is the case with many of the inhibitors currently in clinical trials, an initial lead compound was identified by the time-consuming process of random screening, that is, the testing of large compound libraries for inhibition of protein kinases in vitro.

So this is also not a real answer.

I find one of this year's Science Computational Design of Proteins Targeting the Conserved Stem Region of Influenza Hemagglutinin
very inspiring,http://www.sciencemag.org/content/332/6031/816.full.html