Researchers at AstraZeneca and Uppsala University have published a nice review in the most recent issue of Drug Discovery Today entitled “Limitations and lessons in the use of X-ray structural information in drug design.” Although not focused specifically on fragments, the lessons are broadly applicable. The article presents three assumptions that are made when using crystallographic data:
1) The protein structure is correct.
2) The structure of the ligand and its interactions with the protein are correct
3) The protein–ligand structure is relevant for drug design
“While these assumptions seem perfectly reasonable at first sight,” the authors note, “they are not all necessarily true.” The review then provides examples where these assumptions went awry, as well as the lessons that can be drawn from these mishaps.
The fact that crystallography can be misleading should not come as a surprise to anyone who has worked in drug discovery, but what makes the article worth reading is the range of examples. I should say that I’m a huge fan of crystallography and have found it to be very valuable. Perhaps because of this, it is all the more distressing when it leads me astray. In my experience with fragment-based drug discovery, this has most frequently happened with assumption #3; it is not uncommon to find a lovely fragment making appealing interactions with the protein, but to still have difficulty measuring any affinity, much less improving it.
Anyone have stories or thoughts to share?