ABSTRACT: Nucleosomes are part of the first level of chromatin packaging. They each consist of a histone heterooctamer around which DNA wraps 1.6 times. The histone heterooctomamer is made up of two copies of histones 2A, 2B, 3 and 5. The segment of DNA wrapped around the histones, the so-called "core" fragment, is 147 base pairs long. Neighboring nucleosomes are separated from one another by a stretch of DNA called the "linker," whose size varies depending on organism, cell type, and even chromatin activity. Certain chromatin remodeling factors govern accessibility of DNA to regulatory proteins by repositioning nucleosomes to reveal regulatory sites that would otherwise be occluded by a nucleosome. In contrast to histone modifications such as methylation or acetylation, which are investigated by ChIP-seq, nucleosome positioning data are generated without immunoprecipitation (see Methods below). Instead, micrococcal nuclease is used to digest chromatin to apparent completion, the (well-defined and clearly visible) mononucleosomal core fragment fraction is isolated by gel purification, and one end is then sequenced. Mapping the sequence tag back to the genome reveals the precise position of one end of the core fragment that was protected by the nucleosome; the position of the other end can then simply be inferred by extending the read to a virtual length of 147 bases. Statistical analyses such as occupancy and positioning stringency can then be employed to analyze the local nucleosome landscape anywhere in the (mappable) genome or to infer global parameters of nucleosome organization. In the context of the ENCODE project, nucleosome positioning data are particularly valuable for analysis of the relationship between transcription factor binding, histone modifications, and gene activity. For a general primer on these types of data and analyses, refer to Valouev et al. (2008). For data usage terms and conditions, please refer to http://www.genome.gov/27528022 and http://www.genome.gov/Pages/Research/ENCODE/ENCODEDataReleasePolicyFinal2008.pdf To isolate mononucleosome core DNA fragments from the GM12878 and K562 ENCODE cell lines we followed the micrococcal nuclease (MNase) digestion and isolation protocol as described in Johnson et al. (2006), Valouev et al. (2008), and Valouev et al. (2011) with the following modifications. The precise concentrations of the two flash-frozen cell samples received from the Snyder Lab were not known so, per our standard procedure, we performed a series of digestions titrating the amount of MNase to determine the concentration of MNase for optimal digestion of each sample. Final concentrations of 25 U/µL and 50 U/µL of MNase were used to digest the GM12878 cells and K562 cells respectively at 20°C for 12 min. All other steps in the digestion and isolation protocol were as described. Cells were grown according to the approved ENCODE cell culture protocols (http://hgwdev.cse.ucsc.edu/ENCODE/protocols/cell). K562 and GM12878 were each grown to ~2.5×108 cells. The cells were harvested, frozen and the nucleosome core isolation followed (Valouev et al. 2008). The SOLiD reads were mapped in color-space with the probabilistic mapper, DNAnexus (https://dnanexus.com/). The DNAnexus mapper measures and propagates mapping uncertainty by including both quality values and mismatches in the alignment score calculation. The scores are then scaled across all possible mappings of the read to estimate the posterior probability for alignment to each genomic location. Reads corresponding to posterior probability of correct mapping > 0.9 were reported. Nucleosome density signal maps (bedgraph and bigwig files) were generated by first shifting reads by 74 bp in the 5´ to 3´ direction and counting the total number of reads starting at each genomic coordinate on both strands. These counts are then smoothed using un-normalized kernel density smoothing with a triweight kernel. A bandwidth of 30 bp is used which is equivalent to a smoothing window of 60 bp. The smoothed counts at each position are then divided by the expected number of reads from an equivalent uniform distribution of reads in a ± 30 bp window around that position. If less than 25% of the positions in a ± 30 bp window around a genomic location are uniquely mappable or if the location is part of an assembly gap, the signal value at that position is considered unreliable and not recorded in the signal files. Hence, genomic coordinates that do not have any associated signal value should be considered missing or unreliable data. Genomic coordinates associated with a signal value of 0 are reliably mapable but do not have any signal in the dataset.
