Figure 4

Dinucleotide frequencies affect assessment of false discovery in genomic tiling. a) Sorted scores generated by tiling human-mouse UCSC chromosome 19 alignments (hg17-mm6) in 120 nt windows beginning every 40 nts (for minimum aligned blocks of length 200 nt) (red line). Each window was also scored after shuffling with shuffle-pair.pl (preserves dinucleotides; black line) and shuffle-aln.pl [18] (which does not preserve dinucleotides; blue line). b) Distributions of scores > 0. The left panels show all scores. To minimize the impact of lower "shufflability" (resulting from the additional constraint of preserving dinucleotides) as contributing to higher scores, the plot was also generated limited to sequences where both shuffling methods changed a minimum of 50% nucleotide identities (middle panels), and to sequences where shuffling while conserving dinucleotides exceeded regular shuffling (right panels). Percentages shown in the panels represent the proportion of all tiling windows scoring above the threshold shown.