Genome-wide nucleosome detection based on the dinucleotide position frequencies

It has been discovered that the properties of nucleosome-bound and linker DNA sequences have important effects on nucleosome positioning. On the other hand, the position frequencies of the nucleosome-bound and linker DNA reveal most of their statistical properties. Therefore, two methods based on the statistical properties of the DNA sequences are proposed for nucleosome positioning. The first method defines the score profile based on the position-frequency differences of some dinucleotides which are most different in the nucleosome-bound and linker DNAs. Our second method is defined by combining the differences in dinucleotide position frequencies and the periodicity of nucleosome-bound DNAs. Experiment results on Saccharomyces cerevisiae show that our second method outperforms significantly other algorithms in nucleosome positioning performed in our paper. Furthermore, this algorithm also achieves the highest accuracy and F-score on the Simian virus 40 chromatin even if the dinucleotide position-frequency data are extracted from the S. cerevisiae.