To evaluate whether there was a relationship between G+C content and read depth, we calculated the average coverage of bases within disjoint 100 bp windows across the genome, as well as G+C% also calculated for these windows. Areas expected to have high or low coverage for processing reasons were masked from the analysis, these included the first and last 120b of the reference genome, as well as genomic 100 bp bins that contained repetitive sequences. The coverage was normalised for each run by dividing the coverage in each window by the mean coverage across all windows for that run. A square-root transformation was applied to the run-normalised coverage. After initial inspection, we identified that a number of very large coverage values were the result of an un-masked LSU rRNA in the B. amyloliquefaciens genome. This small region was masked prior to G+C modeling. The relationship between the square-root normalised coverage and G+C% content was evaluated by fitting linear models using the lm function in the R statistical package.
