miRNA-TFBS association compared against four types of negative control sequences. We compared the observed prevalence of TFBS motifs in pre-miR sequences against four types of negative control datasets. For each human pre-miR, 1,000 sequences having the same length as the pre-miR were selected from the human genome according to one of 4 different rules and were examined for the presence of TFBS motifs. In the first negative control dataset, for each pre-miR, 1,000 sequences were chosen randomly in the genome and the average number of sequences containing at least one TFBS motif was scored. In the second negative control dataset, for each pre-miR, negative control sequences were chosen within 5 kb upstream of annotated genes. In the third case, negative control sequences were chosen within introns. In the fourth dataset, negative control sequences were chosen randomly and filtered to retain those which are most similar to pre-miRs in their cross-species conservation and dinucleotide composition (see Methods). Statistical significance was assessed by chi-square test.