Performance of iMBEA and LCM-MBC on GeneWeaver graphs. The GeneWeaver graphs were constructed from two different phenotype-gene similarity matrices. Each edge is either present or absent based on whether it is at or above (or at or below, when p-values are used) a given threshold. The graphs in (a) were created from a correlation matrix of 33 phenotypes and 17539 genes. Graphs in (b) and (c) were created from a matrix of correlation p-values for gene expression to phenotypes in a single mouse population, using 782 phenotypes and 45137 genes. As the threshold moves to the right along the x-axis, the graphs generally grow larger and denser. The pheno-gene graphs range from 6.6% to 34.7% density, while the cerebellum graphs range from about 0.2% to about 2.5% density.