Extended Data Figure 1: Correcting for GC-bias in deep sequencing data.

Sequence composition has previously been reported to influence the depth of sequence coverage³⁴. Therefore we investigated whether GC-content contributes to the noise in our data. Sequence reads from the wild isolate (DS2) stationary-phase sample were analysed with respect to GC-content. a, For each 1 kb window of unique sequence the number of mapped reads was plotted against the GC-content of the window. We found a significant reduction in mapped sequence reads at elevated GC-content. A polynomial equation (inset and solid line) was fitted to the data. b, For each 1 kb window of unique sequence, the read counts were plotted against chromosome position. c, Using the method of Alkan et al.³⁴, we corrected for GC-bias using the polynomial equation shown in a and then plotted the corrected sequence reads against GC-content. d, GC-bias-corrected sequence reads are shown plotted against chromosomal position. With no substantial continuing replication in the stationary-phase sample, we can justify using this data set to normalize the exponential phase data. Both normalization methods result in low noise compared with studies that do not use a normalization step³⁵.