Figure 6

Prediction accuracy on a wide genomic region. Prediction accuracy (ROC) measured for the loci shown in Figure 5. Amount of true positives (TP, green line) is equal to the fraction of correctly predicted DNA positions in CRMs, calculated for different cutoff values. Blue and red lines show the rate of false-negatives (FN) and false-positives (FP) correspondingly. (A) Prediction accuracy based on sequence conservation (PIP) between the two Drosophila species, (B) prediction accuracy based on word frequency analysis, (C) an example of a combined approach; moderate filtering for non-conserved regions (below 30%) allows further improvement of the prediction quality as in (B).