CoChat
Sign Up
Cited by 423Genomics and Chromatin DynamicsBioinformatics and Genomic NetworksCancer-related molecular mechanisms research

The DOI matches its Crossref record; 60 of the 60 references it cites were found in a registry or an index.

CHiCAGO: robust detection of DNA looping interactions in Capture Hi-C data

Open access
Cairns, Jonathan, Freire-Pritchett, Paula, Wingett, Steven W., Várnai, Csilla, Dimond, Andrew, Plagnol, Vincent, Zerbino, Daniel, Schoenfelder, Stefan, Javierre, Biola-Maria, Osborne, Cameron, Fraser, Peter, Spivakov, Mikhail·2016·Genome Biology
DOI 10.1186/s13059-016-0992-2PMC PMC4908757

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Capture Hi-C (CHi-C) is a method for profiling chromosomal interactions involving targeted regions of interest, such as gene promoters, globally and at high resolution. Signal detection in CHi-C data involves a number of statistical challenges that are not observed when using other Hi-C-like techniques. We present a background model and algorithms for normalisation and multiple testing that are specifically adapted to CHi-C experiments. We implement these procedures in CHiCAGO ( http://regulatorygenomicsgroup.org/chicago ), an open-source package for robust interaction detection in CHi-C. We validate CHiCAGO by showing that promoter-interacting regions detected with this method are enriched for regulatory features and disease-associated SNPs.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1186/s13059-016-0992-2

Cites

60 of 60 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 The outline of Capture Hi-C. a Outline of the CHi-C protocol. A Hi-C library is hybridized to a capture system that consists of biotinylated RNA probes targeting the ends of DNA restriction fragments. After hybridization, streptavidin pulldown is performed to filter for fragments that have hybridized with the RNA probes, leading to enrichment in baited fragments (“baits”). Following a limited-cycle PCR amplification, the CHi-C library is ready to be analysed by massively parallel paired-end sequencing. b The interactome of the LPHN2 promoter region in GM12878 cells. The top panel shows a 1.8-Mb region containing the LPHN2 gene. The middle panel shows raw read pairs from the Hi-C library. All read pairs sequenced for these regions are shown in grey. In purple, we show only the read pairs that contain the LPHN2 promoter in one of the fragment ends. The bottom panel shows raw read pairs from the Promoter CHi-C library from [3]. The WashU EpiGenome Browser [58, 59] was used to create this figure
Fig. 1 The outline of Capture Hi-C. a Outline of the CHi-C protocol. A Hi-C library is hybridized to a capture system that consists of biotinylated RNA probes targeting the ends of DNA restriction fragments. After hybridization, streptavidin pulldown is performed to filter for fragments that have hybridized with the RNA probes, leading to enrichment in baited fragments (“baits”). Following a limited-cycle PCR amplification, the CHi-C library is ready to be analysed by massively parallel paired-end sequencing. b The interactome of the LPHN2 promoter region in GM12878 cells. The top panel shows a 1.8-Mb region containing the LPHN2 gene. The middle panel shows raw read pairs from the Hi-C library. All read pairs sequenced for these regions are shown in grey. In purple, we show only the read pairs that contain the LPHN2 promoter in one of the fragment ends. The bottom panel shows raw read pairs from the Promoter CHi-C library from [3]. The WashU EpiGenome Browser [58, 59] was used to create this figureCC BY · PubMed Central
Fig. 2 Examples of interactions called by CHiCAGO. Top panels: plots showing the read counts from bait-other end pairs within 500 kb (upstream and downstream) of two baits, containing the promoters of a VEZF1 and b RGS22 in GM12878 cells. Significant interactions detected by CHiCAGO (score ≥5) are shown in red, and sub-threshold interactions (3 ≤ score < 5) are shown in blue. Triangles indicate bait-to-bait interactions. Grey lines show expected counts and dashed lines the upper bound of the 95 % confidence intervals. (Note that bait-to-bait interactions have higher expected read counts than bait-to-non-bait interactions spanning the same distance.) Bottom panels: the genomic maps of the corresponding regions, with coloured bars showing “chromatin colours” obtained from performing chromatin segmentation with chromHMM [60]: red, active promoter; pink, poised/repressed promoter; orange, strong enhancer; yellow, weak enhancer; blue, insulator
Fig. 2 Examples of interactions called by CHiCAGO. Top panels: plots showing the read counts from bait-other end pairs within 500 kb (upstream and downstream) of two baits, containing the promoters of a VEZF1 and b RGS22 in GM12878 cells. Significant interactions detected by CHiCAGO (score ≥5) are shown in red, and sub-threshold interactions (3 ≤ score < 5) are shown in blue. Triangles indicate bait-to-bait interactions. Grey lines show expected counts and dashed lines the upper bound of the 95 % confidence intervals. (Note that bait-to-bait interactions have higher expected read counts than bait-to-non-bait interactions spanning the same distance.) Bottom panels: the genomic maps of the corresponding regions, with coloured bars showing “chromatin colours” obtained from performing chromatin segmentation with chromHMM [60]: red, active promoter; pink, poised/repressed promoter; orange, strong enhancer; yellow, weak enhancer; blue, insulatorCC BY · PubMed Central
Fig. 3 Read count distributions in typical CHi-C experiments and fit to the CHiCAGO background model. Histograms showing read count distributions for fragment pairs spanning different distance ranges (±20 kb) in a biological replicate of GM12878 (left) and mESC (right) CHi-C data. Solid lines indicate expected count distributions according to the CHiCAGO background model. The expected distribution was derived by taking samples from the null model, then kernel smoothing the resultant histogram to form a curve for visualisation purposes
Fig. 3 Read count distributions in typical CHi-C experiments and fit to the CHiCAGO background model. Histograms showing read count distributions for fragment pairs spanning different distance ranges (±20 kb) in a biological replicate of GM12878 (left) and mESC (right) CHi-C data. Solid lines indicate expected count distributions according to the CHiCAGO background model. The expected distribution was derived by taking samples from the null model, then kernel smoothing the resultant histogram to form a curve for visualisation purposesCC BY · PubMed Central
Fig. 4 Sources of background counts and bias accounted for by the CHiCAGO model, illustrated with GM12878 data. a–c Different factors that the Brownian background component models: a multiplicative bait-specific bias (shown are three representative distance profiles for three different values of the bait-specific bias factor); b distance dependency, plotted on a log–log scale; c multiplicative other-end bias (each bar represents a pool of other ends defined by a range of trans-chromosomal read pairs accumulated by each other end; bait-to-bait interactions are pooled separately). d Technical noise is estimated separately for each combination of bait and other-end pools, with pool membership defined by the number of accumulated trans-chromosomal read pairs. Here, we show the technical noise level estimates for each bait-other end pool combination, grouped together by bait pool
Fig. 4 Sources of background counts and bias accounted for by the CHiCAGO model, illustrated with GM12878 data. a–c Different factors that the Brownian background component models: a multiplicative bait-specific bias (shown are three representative distance profiles for three different values of the bait-specific bias factor); b distance dependency, plotted on a log–log scale; c multiplicative other-end bias (each bar represents a pool of other ends defined by a range of trans-chromosomal read pairs accumulated by each other end; bait-to-bait interactions are pooled separately). d Technical noise is estimated separately for each combination of bait and other-end pools, with pool membership defined by the number of accumulated trans-chromosomal read pairs. Here, we show the technical noise level estimates for each bait-other end pool combination, grouped together by bait poolCC BY · PubMed Central
Fig. 5 CHiCAGO multiple testing approach schematic. a Empirical probability of reproducible interaction (used to generate weight profiles) as a function of interaction distance, generated on two replicates of GM12878 cells, assessed for the 100,000 top-scoring interactions. b–d The effects of applying p value weighting to the GM12878 data. The arrow on the x-axis indicates the number of significant interactions called in the weighted data. Upon applying weighting, we see a decrease in the interaction distance amongst cis-interactions (b). P value weighting increases the mean read count of called interactions (c) and decreases the prevalence of trans-chromosomal interactions (d)
Fig. 5 CHiCAGO multiple testing approach schematic. a Empirical probability of reproducible interaction (used to generate weight profiles) as a function of interaction distance, generated on two replicates of GM12878 cells, assessed for the 100,000 top-scoring interactions. b–d The effects of applying p value weighting to the GM12878 data. The arrow on the x-axis indicates the number of significant interactions called in the weighted data. Upon applying weighting, we see a decrease in the interaction distance amongst cis-interactions (b). P value weighting increases the mean read count of called interactions (c) and decreases the prevalence of trans-chromosomal interactions (d)CC BY · PubMed Central
Fig. 6 Chromatin features of promoter-interacting fragments detected using CHiCAGO. Yellow bars indicate overlaps with cis-interacting fragments within 1 Mb of promoter baits; blue bars indicate expected overlap values based on 100 random subsets of HindIII fragments. These subsets were selected to have a similar distribution of distances from gene promoters as the interacting fragments. a GM12878 CHi-C data. Chromatin features are obtained from the ENCODE project [61]. b mESC CHi-C data. Chromatin features are obtained from the mouse ENCODE project [62]. These plots are generated automatically by the CHiCAGO pipeline. Error bars represent 95 % confidence intervals
Fig. 6 Chromatin features of promoter-interacting fragments detected using CHiCAGO. Yellow bars indicate overlaps with cis-interacting fragments within 1 Mb of promoter baits; blue bars indicate expected overlap values based on 100 random subsets of HindIII fragments. These subsets were selected to have a similar distribution of distances from gene promoters as the interacting fragments. a GM12878 CHi-C data. Chromatin features are obtained from the ENCODE project [61]. b mESC CHi-C data. Chromatin features are obtained from the mouse ENCODE project [62]. These plots are generated automatically by the CHiCAGO pipeline. Error bars represent 95 % confidence intervalsCC BY · PubMed Central
Fig. 7 Significant enrichment for GWAS SNPs at CHiCAGO-detected promoter-interacting regions in human lymphoblastoid cells. Enrichment for SNPs associated with autoimmune diseases (AI), and kidney/liver/lung (KLL) and neurological behaviour (NB) disorders [22] in the CHiCAGO-detected interactions in the GM12878 cell line. The bar plot shows p values for the enrichment of each disorder; red bars indicate p values computed in interacting fragments; blue bars indicate p values computed in 100 random subsets of HindIII fragments selected to have a similar distribution of distances from gene promoters as the interacting fragments. This analysis was performed using the software package GoShifter (Genomic Annotation Shifter) [23]
Fig. 7 Significant enrichment for GWAS SNPs at CHiCAGO-detected promoter-interacting regions in human lymphoblastoid cells. Enrichment for SNPs associated with autoimmune diseases (AI), and kidney/liver/lung (KLL) and neurological behaviour (NB) disorders [22] in the CHiCAGO-detected interactions in the GM12878 cell line. The bar plot shows p values for the enrichment of each disorder; red bars indicate p values computed in interacting fragments; blue bars indicate p values computed in 100 random subsets of HindIII fragments selected to have a similar distribution of distances from gene promoters as the interacting fragments. This analysis was performed using the software package GoShifter (Genomic Annotation Shifter) [23]CC BY · PubMed Central
Fig. 8 Enrichment of promoter-interacting fragments for regions capable of driving transgene expression in mESCs. TRIP (Thousands of Reporters Integrated in Parallel) assesses the influence of local chromatin context on gene expression. This is achieved by integrating a barcoded transgene reporter into thousands of genomic locations in parallel and monitoring the transcriptional activity at each location [24]. Normalised RNA read counts from reporter insertions are separated according to (i) their overlap with HindIII fragments engaging or not in interactions; (ii) their promoter-other end distance. For non-interacting HindIII fragments, distance is measured from the nearest promoter in the linear sequence. Blue and green box plots indicate read count summary statistics for promoter-interacting and non-interacting HindIII fragments, respectively. Each dashed line shows the regression of median log-normalised read counts against promoter-other end distance bin, considering promoter-interacting (blue) and non-interacting (green) HindIII fragments separately
Fig. 8 Enrichment of promoter-interacting fragments for regions capable of driving transgene expression in mESCs. TRIP (Thousands of Reporters Integrated in Parallel) assesses the influence of local chromatin context on gene expression. This is achieved by integrating a barcoded transgene reporter into thousands of genomic locations in parallel and monitoring the transcriptional activity at each location [24]. Normalised RNA read counts from reporter insertions are separated according to (i) their overlap with HindIII fragments engaging or not in interactions; (ii) their promoter-other end distance. For non-interacting HindIII fragments, distance is measured from the nearest promoter in the linear sequence. Blue and green box plots indicate read count summary statistics for promoter-interacting and non-interacting HindIII fragments, respectively. Each dashed line shows the regression of median log-normalised read counts against promoter-other end distance bin, considering promoter-interacting (blue) and non-interacting (green) HindIII fragments separatelyCC BY · PubMed Central
Fig. 9 Circlet view of promoter–promoter interactions for histone genes in GM12878 cells. Interactions where histone gene promoters are engaged at both fragment ends are shown in dark magenta. Interactions where histone gene promoters are interacting with non-histone gene promoters are shown in grey. The WashU EpiGenome Browser [58, 59] was used to create this figure
Fig. 9 Circlet view of promoter–promoter interactions for histone genes in GM12878 cells. Interactions where histone gene promoters are engaged at both fragment ends are shown in dark magenta. Interactions where histone gene promoters are interacting with non-histone gene promoters are shown in grey. The WashU EpiGenome Browser [58, 59] was used to create this figureCC BY · PubMed Central
Fig. 10 Comparison of interactions detected in CHi-C and Hi-C data. a Top panels: plots showing the read counts from bait–other end pairs within 750 kb (upstream and downstream) of three baits, containing the Pax6, Foxo4 and Tbx5 promoters (from left to right). Significant interactions detected by CHiCAGO (score ≥5) are shown in red, and sub-threshold interactions (3 ≤ score < 5) are shown in blue. Bottom panels: raw Hi-C matrices at 25-kb bin resolution within the corresponding 1.5-Mb regions. The bottom corners of the red lines indicate example bin pairs, within which significant interactions were detected in the CHi-C data. b Mapping of short-range (<1 Mb) CHi-C interactions within 25-kb interacting bins detected in the Hi-C data. Filled circles show the observed fraction of CHi-C interactions mapping within the Hi-C interacting bins; open circles show the expected fraction estimated by a permutation strategy accounting for genomic structure (see “Methods” for details). The standard deviations across 100 permutations are not shown as they are smaller than point size. c Mapping of long-range (>1 Mb) CHi-C interactions within 1-Mb interacting bins detected in the Hi-C data. Filled circles show the observed fraction of long-range cis- and trans-chromosomal interactions detected in the CHi-C data that map within the Hi-C interacting bins. Open circles show the expected fraction estimated by a permutation strategy accounting for genomic structure (see “Methods” for details). Error bars show standard deviation across 100 permutations. d The overlap of long-range (>5 Mb) interacting fragment pairs detected in CHi-C data (blue circles) and interacting 1-Mb bin pairs detected in the Hi-C data (black squares) on chromosomes 6 (left) and 11 (centre) and for trans-interactions between these chromosomes (right). All panels present pre-capture mESC Hi-C data from [4]
Fig. 10 Comparison of interactions detected in CHi-C and Hi-C data. a Top panels: plots showing the read counts from bait–other end pairs within 750 kb (upstream and downstream) of three baits, containing the Pax6, Foxo4 and Tbx5 promoters (from left to right). Significant interactions detected by CHiCAGO (score ≥5) are shown in red, and sub-threshold interactions (3 ≤ score < 5) are shown in blue. Bottom panels: raw Hi-C matrices at 25-kb bin resolution within the corresponding 1.5-Mb regions. The bottom corners of the red lines indicate example bin pairs, within which significant interactions were detected in the CHi-C data. b Mapping of short-range (<1 Mb) CHi-C interactions within 25-kb interacting bins detected in the Hi-C data. Filled circles show the observed fraction of CHi-C interactions mapping within the Hi-C interacting bins; open circles show the expected fraction estimated by a permutation strategy accounting for genomic structure (see “Methods” for details). The standard deviations across 100 permutations are not shown as they are smaller than point size. c Mapping of long-range (>1 Mb) CHi-C interactions within 1-Mb interacting bins detected in the Hi-C data. Filled circles show the observed fraction of long-range cis- and trans-chromosomal interactions detected in the CHi-C data that map within the Hi-C interacting bins. Open circles show the expected fraction estimated by a permutation strategy accounting for genomic structure (see “Methods” for details). Error bars show standard deviation across 100 permutations. d The overlap of long-range (>5 Mb) interacting fragment pairs detected in CHi-C data (blue circles) and interacting 1-Mb bin pairs detected in the Hi-C data (black squares) on chromosomes 6 (left) and 11 (centre) and for trans-interactions between these chromosomes (right). All panels present pre-capture mESC Hi-C data from [4]CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 60 references

Research works printed but not linked: 1

  1. NCBI Gene Expression Omnibus. http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE81503 . Accessed 25 May 2016.

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Cairns, J., Freire-Pritchett, P., Wingett, S. W., Várnai, C., Dimond, A., Plagnol, V., Zerbino, D., Schoenfelder, S., Javierre, B.-M., Osborne, C., Fraser, P., & Spivakov, M.
Affiliations
  • Babraham Institute (GB)
  • University College London (GB)
  • European Bioinformatics Institute (GB)
  • King's College London (GB)
Corresponding author
Spivakov, Mikhail
ORCID
Venue
Genome Biology
Publisher
Springer Science and Business Media LLC
Published
2016-06-15
Volume
17
Issue
1
Pages
127-127
ISSN
  • 1465-6906 (ISSN-L)
  • 1474-760X
DOI
10.1186/s13059-016-0992-2
PMC
PMC4908757
OpenAlex
W2950020758
Cited by
417
References
60
Citations by year
2026: 20 · 2025: 34 · 2024: 42 · 2023: 40 · 2022: 45 · 2021: 48 · 2020: 51 · 2019: 52 · 2018: 39 · 2017: 32 · 2016: 11
FWCI
23.22
Type
journal article
Language
en
Open access
Licence
cc-by
Locations
  • Genome biology (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research Portal (King's College London) (submittedVersion, cc-by)
  • University of Birmingham Research Portal (University of Birmingham) (submittedVersion, cc-by)
  • Spiral (Imperial College London) (submittedVersion, other-oa)
Funders
Biotechnology and Biological Sciences Research Council (GB)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Chromatin Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Bioinformatics and Genomic Networks
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Cancer Research › Cancer-related molecular mechanisms research
Keywords
Capture Hi-C, chromatin looping, background modeling, normalization, multiple comparisons, Chicago, promoter interactions, regulatory elements, disease-associated SNPs
MeSH terms
Algorithms, Chromatin, Chromosomes, Humans, Software, Internet, Polymorphism, Single Nucleotide, Genomics

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review