Detecting Motifs from Sequences

The problem of multiple global comparison in families of biological sequences has been wellstudied. Fewer algorithms have been developed for identifying local consensus patterns or motifs in biological sequence. These two important problems have di erent biological constraints and, consequently, di erent computational approaches. The di culty of nding the biologically meaningful motifs results from (1) the variation among motif bases, (2) the alignment of motif position (sites) among the sequences, and (3) the multiplicity of motif occurrences within a given sequence. In this paper, we review and compare the main approaches for nding motifs. We also introduce our own approach, DMS, which combines two objective functions with an improved iterative sampling search method. We demonstrate the e ectiveness of the various algorithms by comparing them on 10 real domains and 14 arti cial domains. The main advantage of DMS is that it is better able to nd shorter motifs.