Fig 1.
Illustrations of how gene networks can be used in cancer data analysis.
A) Information propagation approach: the information about mutated genes (in red) is propagated to their neighborhood through interactions, helping to identify significantly affected subnetworks. The level of redness of a node indicates how likely the gene is affected. B) Module Cover approach finds the minimum cost subnetworks so that each patient is covered by at least k mutated genes. The edges in the gene interaction network (blue edges) may be weighted based on interaction confidence or mutual exclusivity. For example, the patients covered by gene C and D are mutually exclusive. There is an edge between a gene and a patient if the gene is mutated in the patient (black edges). The figure shows an example where two modules are selected, covering each patient at least three times (k = 3). The green nodes are selected genes, and the thick edges indicate the selected interactions or gene-patient relationships.
Fig 2.
Other types of biological networks.
A) Topic model utilizing a patient similarity network. The network guides to find disease subtypes and their features (in the figure, the mutations in genes g1 and g2 are selected features for Subtype 1, while Subtype 2 has mutations in g4 and g5). Patients can be represented as mixtures of multiple subtypes. B) Disease network. A disease network can be constructed based on shared disease genes or the similarity of disease phenotypes. For example, the disease network on the right has an edge between two diseases if they share the same disease genes or phenotype features.