We are pleased to introduce a new, XML based format of the complete dataset file. It conforms to the specification defined in the xin.xsd schema file. However, note, that the format defined there is very flexible and defines only the overall structure of a document that can be used to describe an arbitrary annotated graph. The DIP dataset file is just one particular application of this format.

The overall structure of a XIN document is tripartite. The first section (<attributes>) defines names, types and default values of the attributes that are assigned to all the nodes and edges of the graph. It is followed by a list of nodes and then by a list of edges. The from and to attributes specified for each node refer to the unique id attribute of each node element and thus define the connectivity of the network.

Each element of the graph is annotated with a set of attributes defined in the <attributes> section. The values of the attributes for each node/edge are specified with <att> entries - if not present, the value is set to the default specified in the appropriate <attributes> section entry.

In the particular case of the DIP dataset file, the class attribute assigned to each edge defines the CORE subset of DIP - a set of the most reliable S. cerevisiae interactions identified computationally as described in Deane at al.

In addition to the attributes, each node/edge can be annotated with any number of <feature> entries - currently those are used to specify cross-references to other protein databases as well as are used to list references to the experimental data describing interactions.

Note, that for brevity we only list PMID numbers as the full information about each article can be retrieved directly from PubMed by searching for a particular PMID nuber.

The class attribute specified for each experiment is used to distinguish between small-scale (exp:s) and genome-wide (exp:g) approach whereas the <val> element specifies the experimental method used to detect the interaction.