TopDetection of putative binding sites was performed by scanning the corresponding sequences and comparing to the position specific weight matrices. Click here for more details (see also Stormo and Fields, 1998). The results presented throughout the main text and in this pages correspond to those obtained using a "min_fp" threshold (i.e. the threshold that yields a low number of false negatives while accepting a larger number of false positives in the individual binding sites).
In the case of the human skeletal muscle set, we explored two possibilities for the sequences used for the scanning. In one case (masked), we masked non-conserved sequences (see Single Motifs above). In the second case (unmasked), we used the raw sequences for scanning. It is generally a matter of debate how much conservation there is for transcription regulatory mechanisms. On the one hand, important regulatory signals are expected to be conserved between humans and mice; on the other hand, we expect transcriptional regulation to be more complex in humans (at least for some cases). In the skeletal muscle set that we studied, masking yielded better results (compare masked vs unmasked for random sets of genes and masked vs. unmasked for the skeletal muscle set). However, it is unclear how well this extrapolates for other sets of genes. Therefore, the algorithm allows to turn masking on or off (see algorithm parameters).