Genes with highly similar co-expression relationships were grouped together by performing average linkage hierarchical clustering on the topo- logical overlap dissimilarity measure, and network modules were defined by cutting the clustering tree into branches using a dynamic tree cutting algorithm