Figure 1

Different approaches to genomic signal reconstruction. (A) Reads in a portion of the genome. Positive-strand read starts are shown as spikes above the center line, while negative-strand read starts are below the center line. (B) Extending reads by the mean fragment length in the direction indicated by the strand to which they are mapped. (C) Pile-up curve obtained by adding together the extended read intervals, regardless of direction. (D) Fixed-bandwidth kernel density estimate (KDE) obtained using Gaussian kernels and after shifting each read position by half the mean fragment length in the read direction (positive or negative strand). (E) Adaptive-bandwidth KDE, also obtained using Gaussian kernels. This method creates greater smoothing in areas of low data density and emphasizes enriched regions more strongly.