miRHub analysis

First, we used the seed-based target prediction algorithm TargetScan v.5.2 (Grimson et al., 2007) to determine for each miRNA the number of predicted conserved targets among the genes in our gene sets (genes found upregulated by RNA-seq with fold change > 0 and adjusted p value < 0.1). Each predicted miRNA-gene interaction was assigned a score based on the strength of the seed match, the level of conservation of the target site, and the clustering of target sites within that gene’s 3’ UTR. Finally, for each miRNA, the final targeting score was calculated by summing the scores across all genes and dividing by the number of genes. We repeated this procedure 10,000 times, with a new set of randomly selected mouse genes each time, in order to generate a background distribution of the predicted targeting scores for each miRNA. These score distributions were then used to calculate an empirical p value of the targeting score for each miRNA in our gene set. Genes were selected at random from a pool with similar overall connectivity to the genes in our gene set, and to account for differences in the average 3’ UTR length between the genes of interest and the randomly selected genes in each simulation, the targeting score was normalized by 3’ UTR length.