ABSTRACT
Computational approaches to the genome-wide identifi cation of transcription factor binding sites (TFBSs) have attracted bioinformatics researchers [1-3]. Since the UIPAC matching algorithm was fi rst reported in 1991 [4], the computational methods have evolved to complex multiple-feature prediction frameworks that include phylogenetic footprintings [5], gene expression data analysis [6], Bayesian trees and graphs, and so on [7]. Nevertheless, although
the previous computational approaches to binding site identifi cation have achieved high sensitivity, they have not demonstrated the desired high specifi city comparable to that of experimental methods [8].
