It’s not every day that you see a picomolar inhibitor. This is all the more true for membrane proteins. And fragment-based lead discovery is rarely attempted with membrane proteins. For all these reasons, a new paper by Guang-Fu Yang at Central China Normal University, Jia-Wei Wu at Tsinghua University, and co-workers in J. Am. Chem. Soc. caught my eye.
The researchers were interested in the cytochrome bc1 complex, which is essential for cellular respiration and a validated antifungal target. Starting with the co-crystal structures of molecules such as azoxystrobin bound to the enzyme complex, they computationally replaced the pyrimidine-containing moiety (red in figure below) with a library of 1735 fragments and calculated the binding energies, in a process called pharmacophore-linked fragment virtual screening (PFVS). Several of the top ten hits were synthesized and tested, and all of these had nanomolar potency. Compound 4 was further optimized, again with the aid of computational chemistry, leading ultimately to picomolar inhibitors such as compound 4f.
Those of you of a suspicious nature may be concerned that the methoxyacrylate moiety looks like a PAINfully reactive electrophile. Happily, the researchers were able to obtain a crystal structure of a molecule in this series bound to the cytochrome bc1 complex, showing that the molecule binds non-covalently and in close agreement to the predicted structure.
In some ways PFVS is reminiscent of Silverman’s fragment hopping, another computational screening and linking approach. Such techniques work best when the protein-ligand complex is relatively rigid, making modeling more straightforward than it would be for a more flexible system.
A medicinal chemist could argue that traditional techniques may well have arrived at similarly potent molecules without fragments or fancy modeling. Still, the fact remains that fragments and modeling were used to discover impressively tight-binding compounds, illustrating again the versatility and increasing application of fragment-based techniques.