|
|
|
|
|
Building Physical Network Models
|
In this study, we annotated a network comprised of physical (protein-DNA and protein-protein) interactions by the likelihood each interaction is real, the directionality and effects (activation or inhibition) of each edge, and whether a possible pathway involving the edge is "active". By observing edge attribute values that were directly or indirectly constrained by interaction and expression data, we were able to express potential functions pertaining to protein-DNA interactions, protein-protein interactions and knock-out gene expression data. Details are included in Yeang et al. Ultimately a major goal of this model inference process was to find optimal configurations that maximize a joint potential (likelihood) function over the variables in the physical model (annotated graph) so that probability values reflect the degree of support that each possible physical model has in the available data.
|
|
|
|
|
|
|
|
