Abstract: We have developed a web server, eTFBS, for mining transcription factor binding sites (TFBSs) in human and other organisms. It employs a data-driven pattern mining technique and its kernel algorithm is based on a method we previously developed for mining gapped and ungapped TFBSs in yeast. eTFBS asks for a set of potential target equences and a set of negative sequences from the user, and reports the best ten motifs that are highly overrepresented in the positive set. For the human, mouse, fruit fly and yeast genomes, the input of eTFBS can be a set of gene names or RefSeq (SGD for yeast) identifiers and the server will collect the sequences of the promoter regions specified by the user directly from UCSC Genome Browser. Using three datasets to compare the performances of eTFBS and three current motif finding tools, we find that eTFBS outperforms the other three methods in most cases. It turns out that even with the search region restricted to [-500bp, TSS], eTFBS can achieve a reasonably good performance. Our study suggests that eTFBS is an efficient web server for mining TFBSs in a well-annotated eukaryotic genome.