henhaobio
Sign in

Genome biologyBioinformaticsPeer-reviewed journal article

HiGlass: Web-based Visual Exploration and Analysis of Genome Interaction Maps

Peter KerpedjieviD, Nezar A. AbdennuriD, Fritz LekschasiD, Chuck McCallumiD, Kasper DinklaiD, Hendrik StrobeltiD, Jacob M. LuberiD, Scott Ouellette, Alaleh AzhiriD, Nikhil KumariD, Jeewon Hwang, Soohyun LeeiD, B. AlveriD, Hanspeter PfisteriD, Leonid A. MirnyiD, Peter J. ParkiD, Nils GehlenborgiD

PDFDOIPreprintCodeData

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 3 (4 Jun 2018) · CC BY 4.0

HiGlass, a web-based, containerized viewer, enables continuous multiscale navigation of Hi-C contact maps alongside genomic tracks using composable views linked by location and zoom.

Hi-C features span scales from tens to millions of base pairs. Existing viewers such as Juicebox and Genome Contact Map Explorer lack user-arranged multi-view layouts with selectable locus and zoom synchronization; HiGlass adds composable linked views, viewport projections and server-stored view configurations so Hi-C and epigenomic tracks can be compared across conditions and shared as navigable figures.

Key results

  1. ShownLinked wild-type and ΔNipbl views visually recapitulated published effects of cohesin-loading loss in mouse hepatocytes: TADs vanish in a gene-poor region while A/B compartment patterns persist, and gene-rich regions show stronger, finer compartmentalization.

    Hi-C binned from 1 kb (mm9); near-diagonal enrichment lost across chr14:80 Mb to chr14:100 Mb while the compartment checkerboard stays intact; upstream of chr14:80 Mb, finer A/B division in ΔNipbl; within chr14:50 Mb to chr14:70 Mb, H3K4me3 and H3K27ac correlate more strongly with compartments in ΔNipbl. Qualitative; no statistics reported. · Exploring and comparing different experimental…; Figure legends

  2. ShownOverlaying seven published TAD callers on one synchronized Hi-C map showed little agreement between callers, wide variation in TAD size, and some calls at compartment scale that overlap compartment transitions.

    Eight zoom- and location-linked views (seven caller overlays plus a bare map) on Forcato et al. data; few regions called by more than one caller; across 10 experiments, replicate consistency within a caller did not imply agreement across callers; HiCseg, insulationScore and TADbit give contiguous partitions, while Arrowhead, TADTree, domainCaller and Armatus call discontinuous intervals. · Comparing the results of feature callers; Conclusions; Figure legends

  3. ShownServing tiles on demand kept navigation responsive on the public deployment. Almost all tile requests returned faster than a latency threshold that a controlled user study tied to fewer observations and hypotheses.

    Of 2,333,420 tile requests to http://higlass.io since February 2017, 2,251,251 (> 96.6%) returned in under 0.5 seconds. Hi-C tiles are 256 × 256 matrix slices; 1D tiles cover 1024 base pair segments. · Tiles break down large datasets into manageable…; The HiGlass server fetches data from files and…

  4. SuggestedExploring ΔNipbl hepatocyte data in HiGlass turned up off-diagonal contact patches ('blotches') between short A-compartment regions. These line up with long, multi-exonic, active genes, which hints that transcription and splicing affect how contacts respond to cohesin loss.

    An overview plus two linked detail views on chromosome 14 (mm9) showed patches more enriched in ΔNipbl than in wild type. RNA-seq and ChIP-seq tracks showed that aligned genes (e.g., Dock5, Mycbp2, Gnrh1, Cln5) are virtually always transcribed. No quantitative enrichment statistics are reported. · Exploring and comparing different experimental…; Conclusions; Figure legends

  5. ClaimedIn the authors' feature comparison, HiGlass is the only Hi-C viewer that lets users choose which heatmaps synchronize and whether by location, zoom or both. It is also the only one with viewport projections that show one view's extent inside another.

    GCME is the only other viewer with continuous pan/zoom; the 3D Genome Browser, WUEB, Juicebox.js and Juicebox render fixed discrete zoom levels. HiGlass offers 2D plus horizontal and vertical triangular heatmaps. Logarithmic color scaling is otherwise found only in GCME, and public/private instances only in WUEB. · Feature overview and comparison with other…

Methods, data and reproducibility

A JavaScript client arranges 1D and 2D tracks into views linked by location and/or zoom and requests tiles (256 × 256 matrix slices for Hi-C) from a Python server. The server reads multi-resolution cooler and hitile files whose zoom levels are built by summing adjacent bins at power-of-2 bin sizes; categorical tracks are thinned by an importance value.

  • Schwarzer et al. Hi-C of wild-type and induced ΔNipbl adult primary hepatocytes binned from 1 kb, viewed with H3K4me3 and H3K27ac ChIP-seq and RNA-seq tracks · mouse (mm9 assembly) · GEO GSE93431
  • Rao et al. 2014 Hi-C data with TAD calls from seven callers, generated following the Forcato et al. procedures · n = 10 experiments; 7 TAD callers · GEO GSE63525
  • Tile request logs from the public HiGlass server since February 2017 · n = 2,333,420 tile requests
Methods & tools
HiGlass (JavaScript client, Python server), cooler multi-resolution Hi-C format, bigWig and hitile 1D formats, Docker (higlass-docker container), TAD callers: TADbit, HiCseg, insulationScore, Arrowhead, TADtree, domainCaller, Armatus, Juicebox and Juicebox.js, Genome Contact Map Explorer (GCME), Washington University Epigenome Browser (WUEB), 3D Genome Browser, UCSC Genome Browser (Track Hubs)
Software & environment
JavaScript client and Python server; the higlass-docker container bundles the viewer, the website scaffold and the server. Hi-C must be stored in cooler format; bigWig is converted to hitile. No version pins are given.
Experimental conditions
Hardware is not stated. On the public server, 2,251,251 of 2,333,420 tile requests (> 96.6%) since February 2017 were served in under 0.5 seconds.
Reproducibility gaps
No software versions or hardware are given. The storage overhead of multi-resolution files appears only in Supp. Table 2. The interactive figures and selected public datasets depend on the hosted http://higlass.io instance staying online.
Limitations (6)
  • Storing Hi-C at multiple power-of-2 resolutions in cooler files makes the data larger on disk (quantified only in Supplementary Table 2). The authors accept this cost to limit server-to-client transfer when large regions are viewed. · stated by the authors
  • Before viewing, data must be preprocessed into multi-resolution formats with an imposed chromosome ordering, then loaded into the server by network request or command line. · stated by the authors
  • Finding and integrating relevant public datasets is still hard. Extended server metadata and standardized APIs for querying genomic datasets are future goals and are not yet implemented. · stated by the authors
  • Hi-C maps average distinct dynamic processes over cell populations, and some TAD calls match compartment-scale features. Phenomena attributed to TADs may therefore come from other Hi-C features. · stated by the authors
  • The biological observations (compartment strengthening, ΔNipbl patches, histone-mark correlation) come from visual inspection of selected loci without quantitative tests. They are hypotheses, not established effects. · noted in this summary
  • The comparison with other viewers is a qualitative feature checklist written by the developers. No user study or head-to-head speed benchmark against Juicebox or GCME is reported, and latency is measured only on HiGlass's own server. · noted in this summary

Read this if you need a self-hostable web viewer to compare Hi-C contact maps and genomic tracks across conditions, loci and zoom levels, or to publish interactive figures readers can navigate.

Citations
2,485
FWCI
17.43
Field percentile
top 1%

Abstract

We present HiGlass, an open source visualization tool built on web technologies that provides a rich interface for rapid, multiplex, and multiscale navigation of 2D genomic maps alongside 1D genomic tracks, allowing users to combine various data types, synchronize multiple visualization modalities, and share fully customizable views with others. We demonstrate its utility in exploring different experimental conditions, comparing the results of analyses, and creating interactive snapshots to share with collaborators and the broader public. HiGlass is accessible online at http://higlass.io and is also available as a containerized application that can be run on any platform.

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISJia '19Robinson '18Spies '18Freese '15Cao '18Szot '17
71% semantic matchBioinformatics
Wenlong Jia, Hechen Li, Shiying Li+1 more·bioRxiv·Dec 10, 2019