The researchers screened 1280 fragments against the sodium ion channel acid-sensing ion channel 3 (ASIC3), a potential target to treat pain. Compounds were screened at 1 mM concentration, and 56 of them inhibited >50%. Interestingly, this 4.4% hit rate is comparable to other fragment screens at Merck on completely different target classes. Dose-response curves were generated for the 56 compounds, yielding 32 hits in 12 separate chemical series with ligand efficiencies > 0.3. Similarity searching led to the testing of 250 analogs, but most of these were less active than the original fragments.
In the absence of structural information, two strategies were used to try to improve the potencies. First, a series of 220 analogs were synthesized in which small (approximately 64 mass units) substituents were systematically added at various positions around the fragments. Unfortunately, most of the compounds were inactive, and those that were active were only marginally better than the starting fragment. The second strategy was to make more dramatic changes using parallel chemistry of up to a few hundred compounds. This was more successful: for example, fragment 3 led to compound 24, with 200-fold improved activity, albeit with a loss in ligand efficiency.
The present results are consistent with the analysis that advancing fragment hits to high potency (<100 nM) and/or to a preclinical development milestone in the absence of structural information is unlikely.I think this may be overly pessimistic. While optimizing fragments in the absence of structure is certainly challenging, it is possible. Fragment screening ultimately delivered several distinct series of low micromolar compounds against ASIC3. Had these come from an HTS assay, the job of optimizing them further would not be any easier. Moreover, this report represents another example of using fragment approaches to tackle membrane proteins.