...

In ChIP-Seq, genomic DNA is fragmented and target-protein-bound DNA fragments are purified by immunoprecipitation. These purified fragments are between 100 and 500 base pairs depending on the protocol; however, because ChIP-Seq uses short-read sequencing (25 to 35 base pair reads) to maximize sequencing depth, only the ends of each fragment will be sequenced. Consequently, with single-end sequencing, the forward and reverse strands for the each fragment will be from opposite ends of the fragment. At a protein-binding site, there will be two peaks of read enrichment, one from enrichment of forward strand reads and another from enrichment of reverse strand reads. The average distance between these peaks is termed the effective fragment length. Because the forward and reverse strand peaks are generated from a common set of fragments, the peaks should be roughly symmetrical. By phase shifting the data to the mid-point between the two peaks, a common read density plot can be created that shows single peaks at binding sites. 

...

For paired-end sequencing, Strand Cross-Correlation is calculated from the distribution of distances between the paired reads from the ends of each fragment.  

Let's We will perform Strand Cross-Correlation to identify the effective fragment length we will can use when calling read enrichment peaks. 

...

For the chip sample (blue), we can see the peak at 111 base pairs, corresponding to an effective fragment length of 111 base pairs. This number can be determined by examining the values in the strand_correlation spreadsheet (Figure 2), by moving the cursor over the peak in the graph, or by sorting the data in the spreadsheet. In lower quality ChIP-Seq data, we might also see a peak at the read length. The Strand Separation of Samples graph is also useful as a quality control measure. In lower quality samplesChIP-Seq data, we would also observe a peak at the read length. The ratio between the Pearson correlation coefficient of the effective fragment length peak and the read length peak, normalized with the minimum correlation coefficient, [cc(fragment length) - min(cc)] / [cc(read length) - min(cc)] should be greater than 0.8 to meet the minimum quality standards recommended by the ENCODE project (Landt et al., Genome Research, 2012).

The mock sample (red) does not have an effective fragment length peak because it does not read density peaks to phase shift. It does have a small peak at the sequencing read length of 26 base pairs. 

...

Checking the distribution of reads

BAM files can contain both aligned and unaligned reads. The spreadsheet created during import shows the number of reads that were aligned to the reference genome. A large number of unaligned reads may be the result of poor quality sequencing data or alignment problems. It may also be useful to know how many reads map to more than one location in the genome if the options used during alignment supported multiple-mapped reads. 

...

The titles of columns 2. 0 Single End Alignments Per Read and 3. 1 Single End Alignment Per Read indicate that this is single end data. Column 2 shows the number of unaligned reads, while column 3 shows the number of reads that aligned exactly once. If the BAM files had used in this tutorial included reads that mapped to more than one location in the genome, there would be additional columns. 

...