Datasets used in this study.

The number of modules and the average number of subunits in the modules are listed per dataset, as well as for the nonredundant combination of all datasets (��all��) and of all curated datasets (��all curated��). The SGD pathways and KEGG datasets are curated and consist mainly of metabolic pathways. The PE and socio-affinity datasets both result from clustering Tandem Affinity Purification (TAP) data. The differences between these two datasets include the fact that PE clusters are based on raw data from the study by Krogan et al. [30] as well as from Gavin et al. [14], the similarity score (Purification Enrichment versus Socio-affinity) and the algorithm used to cluster the proteins. MIPS and Aloy are two curated complex datasets, the Aloy dataset is a manual selection based on extensive literature curation, information on protein structures and previous TAP derived protein complexes [26]. Curated datasets comprise approximately one third of all modules.