Figure 4.

Comparison of the degree of completion of the yeast and human protein-interaction networks. Neither (a) the yeast nor (b) the human protein-interaction network is near completion as judged by the extent of assay saturation for the studies indicated here, although the yeast network shows higher saturation. With repeating assays on a finite set of interactions, we expect the rate of discovery of new interactions (gray line) to fall below 100% (black diagonal line) and asymptotically to approach the false-positive discovery rate. If false-positive rates are properly controlled, the rate of new interactions should level out, indicating the complete network assayable by these methods. In yeast, the most recent mass spectrometry study of Krogan et al. [28] (core set) shows 66% new interactions, suggesting initial saturation. Human protein interactions are under-sampled; the most recent study, Rual et al. [14], assayed 95% new interactions.