Figure 1: A: Different forms of protein aggregations lead to protein sets of different similarity in the IP data. All proteins that occur together across experiments at least once can be seen as large set in a very coarse view. Protein complexes occur in a more detailed view and their unique cores in the most detailed view. B: Protein aggregations occcur at different levels of protein similarity. The clusters that are obtained by a clustering method at one level of detail can represent complexes of different types for this reason and it is unclear, what each cluster represents. Our strategy captures complexes at different detail levels for this reason and creates trees that preserve the information at which level a cluster occurs and where it is separated into smaller clusters.