Figure 3

Plot of classification accuracy for the Nucleosome positioning sequences with respect to the gap between the prior and the posterior. This is based on the first order tri-nucleotide model (L ₁ = 3, O = 1, g ₁ = 0, L ₂ = 3, g ₂ = 0) at varying values of G. We achieve the best classification accuracies (peaks) at G = 4, 15 and 25, and worst classification accuracies (valleys) at G = 7 and 18. The distances between consecutive peaks and valleys are around 10–11 bps, which is close to DNA helix turn of 10.5 bps (for most common B-DNA, 11 bps for A-DNA, 12 bps for Z-DNA). This result illustrates the utility of the tool in exploring such long-range dependencies which might indicate specific structural constraints of the sequence class.