Figure 6.

ZINBA calls broader regions of signal and selects sets of peaks that are coherent across datasets. (a) The proportion of the top cumulative sets of MACS (red dashed line), F-Seq (green dashed line) and ZINBA refined (light blue line) RNA Pol II peaks that uniquely overlap a FAIRE-seq peak called by the respective method. For comparison, overlap was also compared using randomly permuted RNA Pol II and FAIRE-seq ZINBA peak calls (black dashed line). (b) The average coverage of the cumulative sets of the top N ranked genes (expression, high to low) by H3K36me3 regions called by MACS (red dashed line), F-Seq (green dashed line) and ZINBA unrefined regions (light blue dashed line). The set of unrefined ZINBA H3K36me3 regions were further clustered throughout the genome to merge nearby peaks (blue solid line) and compared to the ranked list of genes in terms of gene body coverage. (c) Comparison of measured gene expression levels for the set of ZINBA H3K36me3 broad regions that either did or did not overlap a ZINBA RNA Pol II broad region. Those overlapping a ZINBA RNA Pol II broad region had three-fold higher median levels of measured gene expression than H3K36me3 regions that did not have any overlap. (d) Representative view of the set of H3K36me3 broad, FAIRE-seq refined and RNA Pol II refined ZINBA peak calls displayed in the UCSC Genome Browser along with the respective read overlap data. For reference, the set of genes (top row) and RNA-seq data (second row) are included. RPKM, reads per kilobase per million mapped reads.

Rashid et al. Genome Biology 2011 12:R67   doi:10.1186/gb-2011-12-7-r67
Download authors' original image