To predict miR-34a targets, we used miRWalk database, http://www.umm.uni-heidelberg.de/apps/zmf/mirwalk/micrornapredictedtarget.html34, which allows a comparative analysis by eight prediction programs (DIANAmT, miRanda, miRDB, miRWalk, RNAhybrid, PICTAR, RNA22 and Targetscan)35–41. Our search was restricted to miRNA binding sites within the 3’UTR, with a minimum seed length of seven nucleotides between miR-34a sequence and its predicted targets. Due to performance differences between algorithms (i.e. sensitivity and specificity in target prediction), we retained only targets that were predicted by at least four different programs. We then compared this list to the BD GWAS list (140 genes published in16) and found 25 genes overlapping with both lists (see Table 1). For gene ontology analysis of miR-34a predicted targets, we used Targetscan 6.2 (http://www.targetscan.org/) and miR-34 family predicted targets were analyzed using DAVID Bioinformatics Resources 6.7 http://david.abcc.ncifcrf.gov/tools.jsp42. Gene ontology categories and disease enrichment were assayed (see Tables S3–S7).
