Whenever I see the title "Computer-Aided Rational Drug Design" in some paper, my doubts are not about the CADD part but about the R part, that is, how rational has the drug design actually been. More simply put, is it "rational" or "rational in retrospect".
But sometimes, things are simple, and there is obvious rationality in the proportion. Like this J. Med. Chem. study on Indole-3-carbinol, a common dietary constituent from cruciferous vegetables, from which analogs that inhibited Akt kinase were made. I3C apparently forms oligomers in the stomach, depending on the pH. These oligomers are thought to be the dominant active species, although their proportion is only ca. 20%. In the study, the authors decided to study the oligomers depicted below, and tried to find common elements that would help them design better analogs whose proportion naturally would be greater.
The CADD part consisted of simple minimization of the structures and looking at the N-N distance in the oligomers. The hypothesis was that if one constructed other compounds with similar N-N distances, maybe the activity would be the same.
I have to say that in the absence of anything indicating otherwise, I always find such an assumption simplistic. That's because proteins are funny entities, and molecules and active sites are often promiscuous, with similar molecules binding totally differently. This concept has negated many previously accepted hypotheses about "common pharmacophores". People lay things on top of each other, look for common elements, and design new compounds encompassing those elements in the right places. However, nature many times has other plans in her mind, and we get bamboozled when it turns out that these "common elements" bind in quite different places. Also, minimized structures depend on the force field being used, and different FFs give different structures. Lastly, there is no a priori reason for assuming that proteins will bind structures even in local minima, let alone global ones. All this is precisely why de novo ligand design is so tricky.
In this case though, the hypothesis is much more reasonable because the molecules are simple and planar and not very flexible. One may expect them to have a very limited number of conformations. And this is probably what it turned out to be. The authors did some simple modifications indicated below, including rigidifying the linker between the two indole rings, in which they came up with analogs retaining the N-N distance. The analogs worked, further modifications added polar electronegative substituents to enhance potency and selectivity, and everyone was reasonably happy in the end.
It's interesting that they did not see it useful to do some docking and active site examination. Methinks that that could have helped weed out some of the analogs with added polarity but with the "wrong" substituents. In any case, I think this is a good example of how computers can help in the simplest of ways to guide drug design. Sorry...I mean ligand design.
Subrahmanyan Chandrasekhar: A study in fortitude and rigor
1 day ago in The Curious Wavefunction