With a great title like"Chemical Space Travel" I just couldn't pass up this early view article in ChemMedChem. Though I'm not sure that I totally buy into this as a method for discovering new drugs, it is an interesting concept nonetheless. Currently, it is estimated that there are 1020 to 10200 "drugable" organic molecules. As it is impossible sift through all of these structures when searching for new lead compounds, knowing what region of chemical space to explore beforehand might be beneficial. Thus, researchers in the Reymond group at the University of Berne in Switzerland have developed a computer program that serves as a "spaceship" for chemical space travel; a point mutation generator serves as a "propulsion device," and a similarity score serves as a "compass." In simpler terms, starting from any molecular structure "A", this program first completes one of eight possible mutations on each atom/bond in the molecule: atom exchange, atom inversion, atom removal, atom addition, bond saturation, bond unsaturation, bond rearrangement, or aromatic ring addition. Then, the similarity between each mutant and the target compound "B" is measured. The 10 mutants that are most similar to the target "B" and 20 random mutant molecules are carried on for another round of mutation/selection. This continues on until one arrives at the target molecule "B," and along the way thousands of unique structures are generated.
One easy example is illustrated below: Starting from methane, 12 mutations produced cubane--but along the way 6638 unique compounds were generated, taking the 10 most similar to the target (in this case cubane) and 20 random compounds at each mutation step. All compounds that were unstable or not synthetically feasible were eliminated. In the same fashion, from cubane to methanol, there were only 7 steps necessary, and during the process almost 1000 new molecules were generated.
So how could this be used for drug discovery? Well, to do this, the authors investigated the chemical space between AMPA and CNQX (shown below); both are known to be agonists of the AMPA receptor, which is a glutamate receptor in the central nervous system. Using these two compounds, over 559,656 compounds were obtained after after 500 runs, which created this cool looking graph. Colors for the graphs are as follows: AMPA to CNQX, in green; CNQX to AMPA in blue, run-away compounds in gray, AMPA to CNQX mutant series in orange, CNQX to AMPA mutant series in pink, and in red are the best docking compounds--or in other words compounds that actually are predicted to bind into the active site of the AMPA receptor (this was determined through computational docking studies). If you haven't noticed, the novel inhibitor with the best predicted affinity for the AMPA receptor is a combination of an amino acid group from AMPA and an aromatic group originating from CNQX.