performed predictions using all the 187 position weight matrices classified as high quality non-redundant vertebrate (mouse, rat and human) matrices in TRANSFAC Release 12.1 (11). We used the default set of MATCH score thresholds provided by TRANSFAC to allow for 10% false negative results. We also filtered out SNPs in non-conserved TFBS. To find conserved TFBS, we first identified the mouse or rat homolog sequence for each predicted TFBS in the human genome based on 17-way vertebrate multiz alignment from UCSC genome bioinformatics web site. We then ran MATCH on these homolog sequences with the same position weight matrices. We categorized a TFBS as conserved if both mouse and rat homolog sequences also have the same predicted TFBS. Several studies (12–14) show that using both the predicted conserved TFBS together with the regulatory potential score (13) can improve predictions, so we also provide this option on the web server.
