A truer measure of our ignorance

In December 2003, Giot et al. (1) published a systematic investigation of the protein interaction network—the interactome—of Drosophila melanogaster. Giot et al. produced a draft map of 7,048 proteins and 20,405 interactions, which they then refined to “a higher confidence map of 4,679 proteins and 4,780 interactions.” The magnitude of the undertaking led to the study being lauded as the “dawn of systems biology” in a number of commentaries and news releases. Giot et al.'s study was preceded and followed by a number of investigations of the interactomes of other species, ranging from bacteria to humans (2–5). However, none of these studies was able to provide an estimate of the actual size of the interactome being sampled. In a systematic statistical study published in this issue of PNAS, Stumpf et al. (6) provide convincing estimates of the interactome size of four organisms, including humans.

Stumpf et al. (6) estimate that the human interactome comprises ≈25,000 proteins and on the order of 650,000 interactions. These numbers provide a sobering view of where we stand in our cataloging of the human interactome. At present, we have identified <0.3% of all estimated interactions among human proteins. We are indeed at the dawn of systems biology.

The sparse sampling of the human interactome should make researchers distrustful of the numerous studies reporting global analysis of human protein interaction networks. As Stumpf et al. (6) stress, the actual size of the interactome may be one of the only global characteristics that can be estimated in an unbiased manner from small, biased samples. This is particularly true of the human interactome: Although the library of probes in most studies is likely to be unbiased, the set of targets is likely selected on the basis of expectations of …

*E-mail: amaral{at}northwestern.edu