Detection of over-represented motifs corresponding to known TFBSs via motif clustering and matching

Detection of over-represented motifs corresponding to known TFBSs (Transcription Factor Binding Sites) is an important problem in biological sequences analysis. In this paper, a novel motif discovery method based on motif clustering and matching is proposed. Against a precompiled library of motifs described as position weight matrices (PWMs), each L-mer in the data set is matched to a motif base on the match scoreʹs p-value, and then the PWMs are updated and clustered according to their similarity. Motif features are ranked in terms of statistical significance (p-value). We present an implementation of this approach, named MotifCM, which is capable of discovering multiple distinct motifs present in a single data set. We apply our method to the benchmark which has 56 data sets, and demonstrate that the performance of MotifCM on this data set compares well to, and in many cases exceeds, the performance of existing tools.