Fig. 1.

Distinguishability problem of informative and uninformative columns. The figures above utilize a visual representation of the positional nucleotide distribution (Crooks et al., 2004), where the height of each letter is proportional to the nucleotide frequency; in all other figures, the height of the letter is proportional to the nucleotide frequency times the information content of the column. Figures (a) and (b) show an alignment of a pair of columns with identical composition: all ‘A’ in (a) and uniform in (b). Figures (c) and (d) show two competing alignments between the same pair of motifs. These alignments would have the same scores under most similarity scores, including ED, even though one is clearly more significant than the other.