Analysis of Hi-C data using SIP effectively identifies loops in organisms from C. elegans to mammals

Abstract

Chromatin loops are a major component of 3D nuclear organization, visually apparent as intense point-to-point interactions in Hi-C maps. Identification of these loops is a critical part of most Hi-C analyses. However, current methods often miss visually evident CTCF loops in Hi-C data sets from mammals, and they completely fail to identify high intensity loops in other organisms. We present SIP, Significant Interaction Peak caller, and SIPMeta, which are platform independent programs to identify and characterize these loops in a time- and memory-efficient manner. We show that SIP is resistant to noise and sequencing depth, and can be used to detect loops that were previously missed in human cells as well as loops in other organisms. SIPMeta corrects for a common visualization artifact by accounting for Manhattan distance to create average plots of Hi-C and HiChIP data. We then demonstrate that the use of SIP and SIPMeta can lead to biological insights by characterizing the contribution of several transcription factors to CTCF loop stability in human cells. We also annotate loops associated with the SMC component of the dosage compensation complex (DCC) in Caenorhabditis elegans and demonstrate that loop anchors represent bidirectional blocks for symmetrical loop extrusion. This is in contrast to the asymmetrical extrusion until unidirectional blockage by CTCF that is presumed to occur in mammals. Using HiChIP and multiway ligation events, we then show that DCC loops form a network of strong interactions that may contribute to X Chromosome–wide condensation in C. elegans hermaphrodites.

Mustache: multi-scale detection of chromatin loops from Hi-C and Micro-C maps using scale-space representation

“…Chromatin loops are defined as pairs of genomic sites that lie far apart along the linear genome but are brought into spatial proximity by a mechanism called loop extrusion [14][15][16]. Several methods have been developed to detect chromatin loops or statistically significant/enriched chromatin interactions from Hi-C contact maps [9,[17][18][19][20]. These existing methods broadly fall into two groups.…”

“…cLoops uses a permuted local background to estimate statistical significance and can handle a broad range of chromatin conformation capture assays including Hi-C, ChIA-PET, HiChIP, and Trac-loop. While cLoops is, to some extent, scale-free since it utilizes reads at their native resolution, cLoops is very inefficient both in terms of runtime (> 1000× slower compared to SIP [19]) and memory use (> 100 GB per chromosome). Another downside is that cLoops reports loops that are not supported by either HiCCUPS, SIP, or Mustache, whereas these three methods have strong agreement among each other.…”

“…More specifically, HiCCUPS identifies loops by finding “enriched” pixels, that is, locus pairs whose contact counts are significantly higher than that of (1) pixels to its lower-left, (2) pixels to its left and right, (3) pixels above and below, and (4) a doughnut-shaped region surrounding the pixel of interest. While our work was under review, a method called Significant Interaction Peak caller (SIP) was published which also uses image processing techniques to identify chromatin loops in Hi-C data [ 19 ]. SIP applies a set of image processing operations (e.g., Gaussian blurring and contrast enhancement) to pre-process the contact maps and increase the contrast of the potential loops.…”

Cohesin residency determines chromatin loop patterns

“…Clearly, CTCF orientation is not the only determinant of loop position in mammals. Indeed, looping is observed in organisms that do not have the CTCF factor ( Heger et al, 2012 ; Rowley et al, 2020 ). Thus, the fundamental mechanism(s) of positioning loops in most organisms including mammals remains to be determined.…”

Cohesin residency determines chromatin loop patterns

“…Clearly orientation is not the only determinant of loop position in mammals. Indeed, looping is observed in organisms that do not have the CTCF factor (Heger et al, 2012;Rowley et al, 2020). Thus, fundamental mechanism(s) of positioning loops in most organisms including mammals remains to be determined.…”