Table 1 Percentage of known protein-protein interactions in the lists of candidate partners generated by the different co-expression-based approaches. For every protein found in the three analyzed complexes, and represented by at least one EST probe in both datasets, we selected the most representative human and mouse probes. A CLOE analysis with a top 1% cutoff was performed on these sequences. In parallel, the human dataset was ranked for each human EST, and lists of candidates corresponding to the top 1% ranks where obtained (Single organism). The prevalence of ESTs corresponding to other proteins of the same complexes was then determined for both approaches. Finally, to determine the prevalence of correct predictions by the Multiple Species approach, we determined the ratio between the number of links with other proteins of the same complex and the total number of links for all the complex components (data from ).