Association vs causation… While medicine is the domain of knowledge where one really needs to have a good grasp on whether or not two observations in question cause each other or are merely associated with each other, our daily forays in chemistry are filled with these same types of dilemmas, only the stakes are lower. Parsing out association from causation is where real science starts and ends. You might spend a lot of effort chasing after a parameter you think is important because your data analysis suggests that it is, only to find out that there is no causal connection to what really matters in your system.

Frustratingly, sometimes you know that you have no chance of establishing a causal relationship between two experimentally determined parameters, but you hope that there is at least an association. Then you find out that there can never be one because the two events are governed by uncorrelated properties. How frustrating is that? Here is an example that has to do with cyclosporine A, the molecule I mentioned a couple of times on my blog. Below you see its solution state conformation on top and the conformation in its bound state with cyclophilin, which is its cellular target (pdb code 2Z6W). It is easy to see intramolecular hydrogen bonds in the solution conformation. These linkages are the main reason for cyclosporine’s oral bioavailability. You do not see these bonds in the pose that corresponds to cyclosporine’s mode of action, do you? Indeed, all of the polar groups now engage the target. This is a frustrating end of science. You can learn all you want from this molecule and emulate its foldable polar surface area in your favourite macrocycle, yet when it comes to designing analogs to improve bioactivity you will be driven by entirely different considerations. How to avoid these sorts of pitfalls? Phenotypic screens, no doubt.