The scaled similarity predicts known regulatory interactions better than other traditional functional similarity measures. It is also reflective of information about the structural importance of an edge in the regulatory network. (

*a*) Overlap of the top 5000 highest ranking gene-pairs according to the scores given by four different measures of functional similarity. (

*b*) Comparison of the predictive power of

and

compared with other commonly used measures, calculated by determining the maximum

*F*-score of each measure. The expected value for the maximum

*F*-score is shown as a grey dotted line. (

*c*) The harmonic mean of the new shortest path upon edge removal from the experimental regulatory network published by RegulonDB, plotted as a function of several measures of functional similarity. For each measure edges were sorted, with highest scoring edges given the lowest rank. The harmonic mean of the

*informational importance* (new shortest path upon edge removal) for all edges above each rank was calculated and plotted as a function of the edge rank. Blue denotes scaled similarity (

); black, scaled similarity (high

*α*); green, semantic similarity; red, Kappa statistic.

## PubMed Commons