The majority of proteins perform their biological role not in isolation, but in close functional interaction with other proteins. For a systems-level understanding of cellular processes, a complete and accurate inventory of protein-protein interactions is essential. Unfortunately, current experimental data on protein-protein associations are incomplete, scattered over various model organisms, and often error-prone. I will present a data-mining system dedicated to the collection, validation and integration of protein interaction data - including associations predicted solely in silico.

In the second part of my talk, I will describe how this consolidated interaction data is useful for making sense of an exciting new source of data: environmental genomics data. This type of data provides, for the first time, an unbiased view of microbial communities, and we find sets of interacting genes with marked preferences for certain environments - helping us to understand how environmental requirements shape genome content.