Given the variety of techniques available to produce protein-protein interaction data and the large number of studies that are published every day, an enormous effort is required to store this information in a way that is both accessible and intelligible to the user (Figure 7).
Molecular interaction databases aim to fulfil this need by extracting information from scientific publications or, in some cases, by including protein-protein interaction predictions found using computational methods 14.
The storage of interactions in publicly available databases allows access to a large volume of interaction data and subsequent analysis of the interactome.
There are several issues when representing interaction data, including use of nomenclature, representing complex data and cross-referencing to other resources. Different databases have different approaches to solving these issues, and database teams are constantly developing new strategies to improve their representation of the data. They also have different levels of curation, depending on how much detail they capture about each interaction.
We will now discuss how heterogeneous database representation of interactions is, and describe some of the initiatives that aim to reduce this heterogeneity.