Figure 5

Relative di- and mononucleotide counts in genomic, test, and positively-scoring shuffled sequences. a) Dinucleotide frequency enrichments over random sequence (where frequency of each dinucleotide is 1/16) for five genomes, human genomic features, and the ncRNA test set subdivided by species and ncRNA classes. Frequency was computed by counting each of the 16 dinucleotides across the indicated sequences and dividing by the total count of all dinucleotides. The bottom panel indicates enrichment of individual nt over random sequence (where frequency is 1/4). b) Dinucleotide enrichments in high-scoring shuffled ncRNA alignments (i.e. the test set, frequencies derived form top-strand) for five ncRNA search tools. Frequencies were normalized to the ncRNA test set. Score thresholds were selected to maximize (precision × recall) on our test set (threshold values are indicated in parentheses). c) Dinucleotide enrichments among the human strand of 1,000 highest-scoring sequences in shuffled Chromosome 19 genomic alignments (tiled in 120 nt windows every 40 bases).