Many networks are actually an one-mode projection of a bipartite network. For instance, many social networks are derived from a bipartite network of people and groups; a collaboration network is derived from a network of people and products (papers).
Another important fact to note is that information contained in a bipartite network gets destroyed during the one-mode projection. Lehmann2008biclique has a nice illustration.
Carefully chosen null models are important to bipartite network analysis.
- Motifs: Artefacts in statistical analyses of network motifs: general framework and application to metabolic networks
See also http://en.wikipedia.org/wiki/Bipartite_network_projection .
Often, a bipartite network is projected onto one of the node space. There are many ways to do it. One can use simple projection by counting the number of shared neighbors. One can also consider each row or column as a vector and calculate the similarity between the vectors. See Similarity.
- Using networks to measure similarity between genes: association index selection
- Using Random Walks to Generate Associations between Objects
- http://arxiv.org/abs/1403.2933 - Efficiently inferring community structure in bipartite networks. Stochastic block model for Bipartite network
- Community Detection in Bipartite Networks with Stochastic Blockmodels
- Mapping Flows in Bipartite Networks
Some bipartite networks contain information about hierarchical structure in one of the node types. For instance, the hierarchical structure among skills can be inferred from user-skill network.
A package for calculating multiple versions of PageRank-like metrics on bipartite networks.