An interesting paper by Lumb and colleagues recently appeared in J. Med. Chem. By way of background, approximately 15 years ago, Brian Shoichet published a provocative article, in which he presented evidence that promiscuous inhibitors often work by forming aggregates:
http://pubs.acs.org/doi/abs/10.1021/jm010533y
This was a landmark paper because it unveiled a plausible cause of false positives in high throughput screening. The formation of such false positives had nothing to do with any special protein/small molecule interaction but was, instead, a non-specific effect. The community took this to heart and went a bit overboard, as we always do, seeing potential aggregators in all sorts of settings. Lumb and colleagues present evidence that aggregate formation and specific molecular level interaction may not be mutually exclusive. This work provides an example of aggregation-based inhibition of a protein–protein interaction involving tumor necrosis factor α. The reported capacity of small molecule ensembles to mimic protein surfaces is super cool. Unfortunately, the pdb file has not been released yet, so forgive me for not providing my usual pretty picture to accompany a structural biology post. You do have an option to see some nice views of the crystallographically characterized complex in the paper. There are 5 (!) inhibitor molecules that clump together into an aggregate in this crystal, replacing an entire protein subunit. I am not sure that there is a general lesson here, but I love seeing exceptions from the rules.