The data was put through our sequencing pipeline consisting of base calling using the Illumina Pipeline, mapping and alignment with BWA [15], PCR duplication removal with SAMTools [16], and GATK [17] for variants calling.
One way to reduce amplification bias is duplication removal.
We used FASTX-Toolkit to perform trimming, quality filtering, and duplication removal [ 37].
The second update yielded 1727 records (1509 after duplication removal), which were assessed by titles and abstracts.
After duplication removal, 2909 rnpB sequences are unique, 2857 of which belong to bacA, 49 to arcA, 3 to bacB, but none for arcM, see Table 3.
The first update of the basic search resulted in 409 records, which were screened by title and abstract, and duplication removal resulted in 31 new full publications.
To assess the effect of duplication removal, we compared the quality of peaks identified using either all unique mappable reads or distinct reads (reads mapped to same locations were collapsed) with the same threshold (FDR < 0.001).
For example, some studies used all reads to call peaks, while others employed reads after duplication removal thinking that reads mapped to the same location are due to amplification bias [ 12, 29, 34].
Find a duplication removal script.
Reads were cleaned using pyrocleaner [ 38], which applied a low-complexity filter followed by a read-size filter (over 100 bp) and a duplication-removal filter.
