An overview of structure priors. Comparison of edge weights suggested by Heckerman
et al. [65], Imoto
et al. [43] and Bernard
et al. [55]. Rows correspond to prior information. In the left two examples the prior can be described as binary, while on the right it is expressed as a
p-value derived from a second data set. The entries in the table are the weights
κvw for each edge depending on whether
G agrees with
or not. The middle table holds
ξ1 <
ξ2, i.e. edges with motif evidence contribute more than edges without. The structure prior is then a product over the weights for all edges.