An Integrative Multi-Network and Multi-Classifier Approach to Predict Genetic Interactions
Figure 1
One feature is overlaid with another feature to generate an overlay feature. Each feature is treated as an undirected network with the genes as the nodes, and the value of the feature for a gene pair becomes the weight of the edge connecting them. Given two weighted networks and
, corresponding to two features respectively, the value of the overlay feature for two nodes A and B, is the strongest (max of product of weights of the two edges) transitive link, i.e.,
, where
and
or vice versa.