From Santa Fe Institute Events Wiki
Statistical Inference for Complex Networks Workshop, December 3-5, 2008, Santa Fe NM
Joel Bader (homepage)
Searching and clustering multi-modal biological networks
Edges in biological networks represent a multitude of distinct relationships, both physical (protein-protein and protein-DNA interactions; enzymatic transformations) and logical (expression correlation; genetic epistasis). While edges in social networks usually indicate affinity or co-group membership, edges in biological networks often indicate the opposite. We describe algorithms developed to analyze multi-modal networks and present applications to biological data sets. Search algorithms use our recent formulation of parity-dependent graph diffusion kernels for anti-affinity edges (Qi et al. 2008 Genome Research). Clustering algorithms use block-models adapted for heavy-tailed networks, with a Chinese restaurant process for sampling the number of groups. We also discuss improved analysis of networks with conservation laws, with applications to metabolism.