Fig. 1.

Finding relevant interactions as a constraint optimization problem. We seek a set of high-confidence edges present in the interactome that directly or indirectly link the proteins and genes identified in the experimental assays. Because some of the input data may be false positives (arrowhead) or may not be explained by currently known interactome (arrow), our approach does not require that all the input data be connected, but rather uses these data as constraints. Note that the protein product and mRNA transcript of the same gene are represented as separate nodes.