Fine mapping chromatin contacts in capture Hi-C data

Abstract

Hi-C and capture Hi-C (CHi-C) are used to map physical contacts between chromatin regions in cell nuclei using high-throughput sequencing. Analysis typically proceeds considering the evidence for contacts between each possible pair of fragments independent from other pairs. This can produce long runs of fragments which appear to all make contact with the same baited fragment of interest. We hypothesised that these long runs could result from a smaller subset of direct contacts and propose a new method, based on a Bayesian sparse variable selection approach, which attempts to fine map these direct contacts.Our model is conceptually novel, exploiting the spatial pattern of counts in CHi-C data, and prioritises fragments with biological properties that would be expected of true contacts. For bait fragments corresponding to gene promoters, we identify contact fragments with active chromatin and contacts that correspond to edges found in previously defined enhancer-target networks; conversely, for intergenic bait fragments, we identify contact fragments corresponding to promoters for genes expressed in that cell type. We show that long runs of apparently co-contacting fragments can typically be explained using a subset of direct contacts consisting of < 10% of the number in the full run, suggesting that greater resolution can be extracted from existing datasets. Our results appear largely complementary to the those from a per-fragment analytical approach, suggesting that they provide an additional level of interpretation that may be used to increase resolution for mapping direct contacts in CHi-C experiments.

Chromatin interactome mapping at 139 independent breast cancer risk signals

“…We then compared the CHiCAGO and MPPC scores for each bait-prey pair. As reported by Eijsbouts et al 30 , we noted that CHiCAGO and MPPC scores were positively correlated ( Figure S3E; Spearman's r = 0.22-0.37). Peaky was able to refine the number of CHiCAGO-scored interactions by 12-17% in both captures; however a proportion of interactions were identified by Peaky but not CHiCAGO (Figure S3F).…”

“…For the VCHi-C, ~11% of CCV-containing fragments interacted with an annotated protein-or non-coding promoter and for the PCHi-C, ~2.5% of promoter fragments specifically interacted with a CCVcontaining fragment ( Figure S3B and Table S4). There were fewer interactions detected by Peaky, perhaps because Peaky can distinguish and rank a subset of direct contacts from long stretches of chromatin interactions 30 . The median linear distance between interactions from either capture was longer than CHiCAGO-scored interactions (ranged from 294-489 kb; Figure S3C).…”

“…It is hypothesized that such collateral contacts might result from inaccuracy during the cross-linking process in CHi-C 28 or from bait migration via Brownian motion 29 . Therefore, as a complementary interaction scoring method, we also used a recently developed Bayesian sparse variable selection approach ("Peaky"; 30 ). The model proposes that for any given bait, the expected CHi-C signal at each prey fragment is expressed as a sum of contributions from a set of fragments directly contacting that bait 30 .…”

“…Therefore, as a complementary interaction scoring method, we also used a recently developed Bayesian sparse variable selection approach ("Peaky"; 30 ). The model proposes that for any given bait, the expected CHi-C signal at each prey fragment is expressed as a sum of contributions from a set of fragments directly contacting that bait 30 . We applied Peaky to the ~1300 baits from the VCHi-C and ~3200 baits from the PCHi-C (Table S4) to derive a measure of confidence in the location of a direct contact called the marginal posterior probability of a contact (MPPC) 30 .…”

Chicdiff: a computational pipeline for detecting differential chromosomal interactions in Capture Hi-C data

“…CHi-C data are often sparse, particularly at large interaction distances, limiting the power of differential signal detection at single-fragment resolution even at significantly interacting regions. In part, this problem can be mitigated based on the fact CHi-C signals commonly spread to adjacent fragments (Eijsbouts et al , 2018) , most likely owing to the tethering of these fragments into the vicinity of the baits by nearby specific interactions. Therefore, to increase power, Chicdiff pools reads across several fragments (by default, five in each direction) surrounding each interacting fragment of interest for each bait.…”