Fig 1.
MAPS workflow contains three major components: pre-processing, normalization and long-range interaction determination. Valid read pairs are obtained after pre-processing and then grouped to short-range reads or long-range reads depending on the genomic distance between two ends (1Kb is used). The long-range reads are further classified into the “AND”, “XOR” or “NOT” set. MAPS only considers the “AND” and “XOR” set of bin pairs for normalization and identification of the statistically significant long-range chromatin interactions. Short-range reads are used to estimate and correct for biases introduced by the ChIP procedure.
Fig 2.
Comparison of sensitivity of MAPS and hichipper.
The Y-axis is the sensitivity, defined as the percentage of detectable HiCCUPS loops of deeply sequenced in situ Hi-C data (S6 Table) recovered by MAPS- or hichipper-identified interactions.
Fig 3.
CTCF motif orientation of MAPS- and hichipper-identified interactions.
The proportion of convergent, tandem and divergent CTCF motif pairs among testable MAPS- and hichipper-identified interactions. Only interactions with both ends containing either single CTCF motif or multiple CTCF motifs in the same direction are considered. The dotted vertical line indicates the expected convergent proportion from randomly chosen CTCF motif pairs (25%).
Fig 4.
Genome-browser shows MAPS-identified interactions anchored at Mtnr1a promoter from mESC H3K4me3 PLAC-seq data.
Anchor regions around Mtnr1a promoter are highlighted by yellow box (chr8:45,065,000–45,075,000, two 5Kb bins). The MAPS-identified interactions overlapping this anchor region are marked by magenta arcs. The black arrow points to the interaction verified in the previous publication [20] and the other end of the interaction is marked by magenta box. Additional interacting regions identified by MAPS are marked by grey boxes. No interaction is identified by hichipper anchored at this region from mESC H3K4me3 PLAC-seq data.
Fig 5.
Cis-regulatory elements are enriched in the target bins of MAPS-identified “XOR” interactions.
As only interactions from the “XOR” set are considered, CTCF enrichment analysis is not applicable for mESC CTCF PLAC-seq data, H3K4me3 enrichment analysis is not applicable for mESC H3K4me3 PLAC-seq data, and H3K27ac enrichment analysis is not applicable for GM12878 H3K27ac HiChIP data (denoted as N.A. in the heatmap). For each ChIP-seq/ATAC-seq data, we calculated the proportion of target bins and control bins containing ChIP-seq/ATAC-seq peaks, defined as %target and %control, respectively. We further defined the enrichment score as the ratio between %target and %control (numbers in S9 Table).
Fig 6.
Enrichment of CREs around the target bins of XOR set of MAPS-specific interactions.
Enrichment of H3K27ac (ChIP-seq peaks), H3K4me1 (ChIP-seq peaks), ATAC-seq peaks, H3K4me3 (ChIP-seq peaks) and CTCF (ChIP-seq peaks) in a window of 500Kb around the target bins for all four datasets. Due to the definition of XOR set of interactions, H3K27ac, H3K4me3 and CTCF enrichment level is not analyzed for GM12878 H3K27ac HiChIP, mESC H3K4me3 and mESC CTCF PLAC-seq data, respectively.
Fig 7.
Frequency of MAPS-identified interactions and control bin pairs versus their rankings in the SPRITE contact matrix (see Methods for details).
A bin pair with higher normalized SPRITE interaction frequency tends to rank top, among all bin pairs with the same genomic distance (only bin pairs in “AND” and “XOR” sets from the SPRITE contact matrix are considered).
Fig 8.
MAPS-identified interactions from mESC H3K4me3 PLAC-seq data anchored at Med13l promoter.
Anchor region around target promoter is highlighted by yellow box. The MAPS-identified interactions overlapping this anchor region are marked by magenta arcs. The deleted enhancer region in Moorthy et al study [24] is marked by magenta box.