Figure 1

The DART algorithm. Flowchart describing the steps in DART (Denoising algorithm based on Relevance network Topology). (1) A synthetic perturbation mRNA signature (e.g derived from an in-vitro cell model, or from a curated list) is evaluated across clinical tumour specimens of a given cancer where pathway activity estimates are sought. Only a subset of genes will show correlations in response to differential pathway activation across tumours. (2) Construction of relevance correlation pruned network as the maximally connected component where all edges reflect correlations that are consistent with the prior information given by the synthetic perturbation signature. (3) Pathway activation over pruned network via a topology based metric, which gives more weight to the hubs in the network. The hypothesis is that correlation hubs represent corresponding pathway activity more faithfully than non-hubs.