Protein interaction: same network, different hubs

Recently, large-scale experiments have provided new insights into the complex protein interaction network in yeast. However, previous analyses have shown that the number of interacting pairs that are common to different methods is extremely low and, therefore, less informative than expected. In this article, we show that comparing the connectivities of individual proteins can reveal that a common tendency between methods has been missed by the pairwise comparison of interactions. We found significant correlations between experimental methods and also between various in silico methods. Exceptionally, a computational method, gene neighbourhood, correlates with both in silico and experimental approaches.