Figure 4.

Illustration of how variability in an INPUT-seq profile can affect reconstruction of average signal profile at TSS and TES. The top panel shows the average signal profiles at the TSS and TES for the ChIP-chip and ChIP-seq profiles of H3K27Me3 at E-16-20 h. These ChIP-chip and ChIP-seq profiles differ quite substantially, and the ChIP-seq profiles resemble that of the GC content variation (Figure 1c). We subsequently reprocessed the ChIP-seq sample by using the INPUT-seq at AdultFemale as background for normalization since this profile has a strong correlation with GC content variation, which more likely reflect the actual technology-specific biases of our sequencing platform. After this procedure, the average signal profiles of ChIP-chip and ChIP-seq look much more alike, indicating that the original INPUT-seq at E-16-20 h does not appropriately capture the technology-specific variation at these sites.