Prioritization

Our method performs an information propagation across an heterogeneous network in two steps: first, it propagates the information within a network using a method based on the Flow Propagation algorithm, using a normalized matrix to propagate labels across nodes in a network. This approach is very similar to a Random Walk with Restarts, with the difference of the normalization process made before. Then, after the propagation within a network, our method propagates the information across different networks using the nodes of the network that are connected to the next network and their label values after within-network propagation. This process is perform for each network in the path from the query network to the target network, and also for each path connecting the query network to the target network. After performing this two-step propagation, a correlation score is computed, comparing the label results of propagating query nodes from query network to the target network, and the label results obtained propagating the target nodes within the target network. This method has been proven useful in past approaches.

For a more detailed description on how ProphTools propagates information, please refer to our publication in gene-domain-disease prioritization.