Sequence logo and mutual information.

Figure a) depicts the sequence logo of CTCF binding sites predicted by the optimal PMM model. We find a high similarity to the previously known CTCF sequence logo [6]. Figure b) depicts the MI of different order between adjacent positions. MI values with a -value above are considered to be insignificant and displayed by the symbol. All MI values of first, second, and third order are significant, and the MI values of fourth order show significance only at some positions. We find that the amount of statistical dependencies varies within the motif to a great extent. We observe high and significant MI values at positions that are comparatively unconserved in the sequence logo, most notably at positions 17 and 19. At very conserved positions, e.g. position 13, the MI value is very low.