Study of fuzzy resemblance measures for DNA motifs

TFBSs are known as regulatory motifs and can be represented as position frequency matrices (PFMs). The de novo identification of transcription factor binding sites (TFBSs) is a crucial problem in computational biology and includes the issue of comparing putative TFBSs to one another and to already known TFBSs. To date there is no fuzzy approach for this problem. In this work we propose the use of fuzzy measures to deal with motif comparison tasks. We investigate the behavior of different classes of classical measures for fuzzy sets including set-theoretic (Jaccard's method), proximity-based (Minkowsky's r-metric), angular coefficient-based (Bhattacharyya's distance) and a measure defined for the fuzzy polynucleotide space. We show that fuzzy measures provide excellent results when dealing with sets of randomly generated motifs and outperforms other existing measures when facing datasets of real motifs.