Fig. 2.

A graphical representation of the workflow used to identify microRNA (miRNA) target sites and polymorphisms in a microRNA target site. The 103,949 possible “major allele” target sites and the 105,466 possible “minor allele” target sites identified by TargetScan, miRANDA, and TargetSpy were filtered (using a “basic score filter”) to produce a list of high-confidence target sites. We then compared major and minor allele-bearing sequences to identify target sites affected by the presence of single-nucleotide polymorphisms (SNPs). Target sites affected in the 3′ supplementary region were subjected to the “score difference filter.” For all sites affected within the seed region and those having passed the “score difference filter,” we compared the results produced by TargetScan, miRANDA, and TargetSpy. Only sites predicted by at least two of the algorithms (the “multiple algorithm filter”) were selected. The selected sites were filtered on the basis of association between the SNP and Alzheimer’s disease (AD). We next assessed other miRNA target sites possibly affected by the four identified SNPs (no basic score filtering + SNP filter). We again applied the multiple algorithm filter. Only one of these miRNAs was also known to be deregulated in the AD brain (the “miRNA AD expression filter”)