C-State (Chromatin-State) was conceived to address the needs of a growing user base in our lab that works with large NGS datasets. Despite the existence of several tools for epigenetic analysis, identification of common epigenetic patterns and their visualization across gene subsets in the context of expression profiles remains a difficult task for biologists. C-State is an interactive platform that enables experimentalists who lack bioinformatics support to investigate epigenetic landscapes across multiple datasets for comparative epigenomic analysis.
The first version of C-State (v0.1) was released in April, 2016. Several new features have since been added based on user inputs. You can use the "Launch C-State" button on the right to open the latest stable version (0.2.3) or you can try the development version (0.2.4-dev). The source code of C-State is available at our GitHub page under an MIT license.
Current version: 0.2.3
OS: Windows, MacOS, Linux
Browsers: Tested on Chrome, Firefox, and Edge
C-State identifies epigenetic patterns across genomic regions and filters genes based on their chromatin signatures. It takes advantage of the “small multiples” paradigm (Tufte, 1990) to facilitate simultaneous visualization and comparison of epigenetic marks on genes in the context of their expression profiles. C-State follows an MVVM (Model-View-View Model) architecture, and is written in JavaScript (ES2015) using VueJS (View layer) and d3.js (plotting). Its features include:
C-State is an integrated visualization platform that enables automatic retrieval, processing and querying of data from whole genome experiments (ChIP-chip, ChIP-Seq, microarray, and RNA-Seq). Using C-State, it becomes easy for bench biologists to accomplish the following tasks without resorting to command-line tools:
C-State automatically extracts data for the target genes in the list, along with the chosen flanks, from whole genome ChIP-seq / ChIP-chip datasets (see steps 4 and 5).
Note: If you have used C-State before and want to re-analyze your data, you can directly load the previously saved C-State JSON file instead of the gene list and select “Previous Plot Data” as the File Type.
C-State currently supports feature information in BED and BED-like formats such as broadPeak and narrowPeak. For a detailed description of these file formats, please see this page.
The View accordion opens to display feature plots for all the genes of interest across the loaded cell types.
C-State uses the small multiples paradigm to render simultaneous plots for comparison. The default grid view is organized for simultaneous depiction of all peak features at the target loci across multiple cell types; data panels are arranged column-wise and each gene is displayed side by side across all samples, facilitating comparative visualization. C-State uses dynamic width for plots and adapts to the number of cell types / conditions specified so that the plots are not rendered off-screen. Scrolling through the grid layout (drag scroll enabled) allows rapid browsing through all the genes of interest.
5 buttons on the left side of the Files and View accordions provide access to other functionalities of C-State.
One of the novel advantages of C-State is that it enables filtering and searching for cell type specific epigenetic features and patterns. This is particularly useful for biologists to identify gene or cell type specific epigenetic patterns without any need for programming or bioinformatics.
The filters module (1st button in the control panel) allows the user to search for genes containing specified patterns of genomic and/or epigenetic features. Options in the filters appear automatically in the drop down menus based on the raw data files provided to C-State. The filters can be chained together in the order of their selection to build complex user-defined pattern searches since each output gets applied to the successive filters in the chain on clicking the “Apply Filters” button (see use case examples).
Search for a gene by name or ID; a set of gene names can be also be pasted to display all the genes that match the given name(s). Ticking the 'Allow partial match’ or 'Match beginning only' checkboxes allows the user to filter for all the genes which have similar names (eg., gene families such as Hox, PAX etc).
Specify size cutoffs for genes or the total regions (including flanks) using a range of operators provided in the drop down.
Search for genes/transcripts that fall within a given expression range in a cell type.
Show or hide genes belonging to a particular chromosome. Chaining multiple chromosome filters allows viewing genes across any chromosome combinations or across all chromosomes. This filter can also be used to display precise co-ordinates of a locus.
Search for genes according to the presence of neighboring genes around them using a range of operators. User can optionally choose to ignore or filter the dataset based on presence of other genes overlapping the target genes.
Apart from the above genomic filters that search based on gene size, location or genomic context, C-State also provides options to identify genes based on their epigenetic patterns.
Select from a set of operators to find genes that carry the desired number of cell type specific marks. The search can be further refined by specifying the distance of the marks from the TSS.
Filter to display genes based on cell type specific patterns of epigenetic marks. Select the two features that are to be viewed in the context of each other from the two drop-down menus and define their relationship using the relation dropdown:
The distance allowed between the two marks can also be specified; in case of overlaps, the min and max distance allowed between the two marks is 0 and this option gets disabled. The search can be further refined by specifying the distance of the pattern from the TSS.
Once the filters are applied using the "Apply Filters" button, the filter panel slides back. The number of active filters are indicated in the header of the View accordion. Filtered output can be exported and saved.
In addition to gene centric plots for visualization, C-State also allows users to identify cell type specific global trends in their dataset using the Plots and Analysis button (2nd button in the control panel). If filters are activated (from the previous panel), the plots can be generated only on the filtered gene subsets thus enabling a platform to analyze multiple specific versus global patterns.
Display the frequency of features based on their size or score (X-axis) across all marks and cell types (arranged row- and column-wise respectively by default). Use the "Switch Rows/Columns" button to toggle the arrangement of cell types and features.
If "Features Scores" is chosen in the dropdown menu and cell types are arranged column-wise:
If rows and columns are switched, the same steps are followed, except that the minimum and maximum values are calculated using the info of respective features across all cell types.
Plot the distribution of features averaged across all genes (Y-axis) along the region specified by the user (X-axis). The gene start and end positions and the chosen upstream and downstream flank sizes are indicated. Data can be plotted for a filtered subset (if any filters are activated) or for the entire list of genes and the Y-axis scale can be adjusted from the dropdowns above the plots.
All genes and their respective flanks are divided into equal number of bins as follows:
Plot the relative gene expression values between pairs of cell types. Expression level of a gene in the first cell type (column name) is plotted on X-axis, and its expression in the second cell type (row name) is plotted on Y-axis.
Tabular information (3rd button in the control panel) of the gene set is provided to display gene information of all or filtered subsets of genes (if filters are activated). The table is interactive and can be sorted on any of the columns. Clicking on the gene name opens the gene modal view for that gene. This feature is useful for sorting information based on user-defined criteria and subsequently viewing the gene panels in the desired order.
The entries can be searched based on terms in any of the columns and arranged page wise using the controls at the top of the table. Clicking the “Copy to clipboard” option copies all the gene names displayed in the table for handy pasting into any other application or relevant rows may directly be copied from the table. Selected pages or the entire table can be exported and saved.
The Downloads button (4th button in the control panel) provides 3 options to save and export the various outputs generated in C-State.
Export C-State Summary - Export a txt file containing summary of the gene names, feature track settings, and active filters (if any).
Download View panels as SVG - Download all the gene panels in the View accordion as a single SVG file.
Download C-State JSON File - Download the entire session info and plot data as a JSON file. This file can be uploaded as "Previous C-State Data" from the Files accordion.
Settings changed from this panel apply across C-State, for visualization, filtering, plotting and analysis and also get saved in any exported / downloaded files.
Feature track settings can be changed to define the quality of the peaks to be displayed (based on feature size and/or score). The user can provide cut-offs to ensure that only peaks meeting the selected criteria are displayed. The color of each of the feature tracks can be selected from 5 color schemes.
Feature bar attributes (height, color) can be defined by the user. By default, exons are not displayed in the main view gene panels but they can be set for display here, along with neighboring genes. The range (default is 5th and 95th percentile) and color palette (default grayscale) used to represent the gene expression level can be customized.
Similar settings as in the main view are available here. These can be changed independently or linked to the settings in the main view.
| Component | Color | Hex Value | RGB Value |
|---|---|---|---|
| Gene Bar | Black | 333333 | 51,51,51 (Hue = 0) |
| Region Bar | Steel Blue | 4682B4 | 70,130,180 |
| Neighbors | Silver | C0C0C0 | 192,192,192 |
Addressing biologically relevant questions often involves analyzing sets of genes belonging to particular pathways or regulating distinct cellular processes. However, extracting chromatin peak information of selected genes of interest from genome-wide datasets is a cumbersome task. The following use cases demonstrate the utility of C-State in the analysis of 16 epigenetic (4 epigenetic marks across 4 different cell types) and 4 RNA-Seq datasets from the ENCODE project. We have focussed on data from multiple human cell lines – K562, HeLa, and GM12878 – for comparison with H1 embryonic stem cells (H1-hESC) to examine changes in histone modification profiles. Whole genome ChIP-seq datasets are downloaded for H3K4me3, H3K9ac (associated with gene activation), H3K36me3 (active transcription) and H3K27me3 (repression). These datasets (see sample format) are loaded directly into C-State as feature files (Tutorial). The FPKM values of all genes derived from RNA-Seq datasets of these cell types are loaded as expression data files (see sample format). Gene expression pattern along with associated histone marks are analyzed across a selected set of 330 target genes and 20 KB of their flanking upstream and downstream regions.
To identify enrichment patterns at a selected subset of genes in these differentiated versus pluripotent cell states, we created a list of 'stemness' genes potentially important for regulating the ES cell state from published datasets analyzing the hESC transcriptome (Bhattacharya et al., 2004) and pluripotency factor bound gene networks in hESCs (Boyer et al., 2005). A subset of 330 genes, shortlisted based on their change in expression profile upon ESC differentiation, is used here for comparative analysis.
The chained filtering application of C-State (Filters panel) allows identification of genes bearing complex cell type specific epigenetic patterns as exemplified by the following searches
Search for bivalent promoters in ESCs
Genes that have bivalent promoters (marked with both
H3K27me3 and H3K4me3 within -5KB to +2KB of TSS) in ESCs can be identified using the "Feature Overlaps" Filter
(image below on the left) chained to a couple of "Feature Counts" Filters set for the absence of the other marks
(image below on the right).
This returns just 13 (of 330) genes. Their individual gene modals can be examined from the View accordion or a list of details obtained from the Tables panel ("Show Filtered Genes only" box checked). The bivalent gene names can also be directly copied to the clipboard for use in other applications.
Note: The images of all the bivalent genes and their individual modals, summary of applied filters, tabular list and plots can be saved.
To further identify genes where the promoter bivalents resolve into a repressed chromatin state in one cell type (GM12878, left image below) and an actively marked one in another (K562, right image below), simply add the appropriate filters to the chain.
Applying this chain of 7 filters returns the muscle specific gene Desmin (DES), which has a bivalently marked promoter in ESCs that resolves into distinct chromatin states in other cell types.
Note: Peak score cutoffs can be applied from the Settings panel to restrict the search to features with a good score.
Active Transcription in ESCs
Genes that are H3K36me3 enriched within 500 bp near exons in
ESCs, indicative of active transcription, can be identified using "Feature Overlaps" Filter set for a maximum distance
of 0.5 KB between a H3K36me3 peak and an exon.
The filter returns 97 genes that match the above criteria. To further identify genes with high transcript levels, add an "Expression" filter to the chain. Set the cell type as H1-hESC, and set the minimum expression value to 3.2 (95th percentile of the loaded datasets, as depicted by the legend in the main view). This return a list of 19 genes. Analyzing these genes using the Average Feature Profile plots shows that they are generally devoid of H3K27me3 mark at their promoter regions. This indicates that the maximal gene expression in H1-hESC cells is dependent on both the presence of H3K36me3 at exons and the absence of H3K27me3 at the promoters.
Analyze Gene Families
Compare the status of the paired box (PAX) family of transcription factors
using "Name" Filter with "Match beginning only" checkbox ticked.
PAX genes code for developmentally important tissue specific transcription factors needed for lineage specification. The 3 PAX genes returned by the filter show a different epigenetic profile at the TSS in ESCs compared to the differentiated cell lines (figure of View accordion below). In ESCs alone, they appear to be actively (H3K9ac) or bivalently marked (H3K4me3/H3K27me3) while the other cell types are associated with the repressive (H3K27me3) mark.
Analyze global peak trends across TSS
The distribution of enriched peaks averaged across all genes can be viewed in the context of their position with respect to the gene body (TSS to TES) using the Average Feature Profile plot (Plots and Analysis panel). For the current test data of 330 genes, the plots indicate an enrichment of H3K27me3 around the TSS in ESCs but not in other cell types. The distribution of other marks, however, remains the same across all cell types. This suggests that, along with the active marks, these genes are also marked with the repressive histone mark at the TSS specifically in ESCs.
Selecting the option to display TSS only (with 20KB flanks) allows investigating this pattern more closely; Filtering the genes for bivalently marked promoters in ESCs reveals a bimodal distribution of H3K27me3 peaks around the TSS - a pattern unique to ESCs.
Search for non-coding genes using the Tables panel
The table view lists information in multiple columns for the genes in the gene list. Sorting for CDS size arranges the genes in ascending order of CDS size. As shown below, the first 10 genes have a CDS size of 0. These may be non coding genes and the gene modals for these can directly be opened (click on the gene name in the table) to view the peak data for this subset across cell types.
This list of non-coding genes can be either exported to an Excel file, or the gene names can be copied to clipboard using the buttons on the bottom right.
C-State can also serve as a simple way to share data and basic analysis of NGS datasets. Following visualization and analysis, the user can save their settings and download the json file of the session from the Downloads panel. This C-State file can then be shared or mailed to their collaborators without needing to share the raw data or any other files. The json file simply needs to be uploaded as Previous Plot Data file into C-State and retains full functionality for visualization and downstream analysis at the other end.
In this section, we have demonstrated some of the obvious use cases of C-State. Users can also visualize other datasets such as different experimental conditions/treatments, cancerous vs normal tissues, transcription factor and other protein binding sites, genomic features such as restriction sites, CpG islands, DNaseI hyper-sensitive sites etc.
Note: The recommended way to use C-State is via the 'Launch C-State' button at the top of this page. Local installation is only needed when C-State is to be hosted on an internal network or a system not connected to the internet. To host C-State locally, follow the steps below. Some technical proficiency is expected.
As web browsers do not have permission to access the local filesystem, C-State needs to be hosted as a static server so that it can access its internal files.
To host C-State:
Open a Terminal/Command Prompt window in the extracted folder:
Windows: Open the C-State folder, go to the explorer bar on the top, type "cmd" and press Enter
Mac: Open a Terminal window (either from Utilities, or by searching in Finder). To browse to the C-State folder, type cd and drag-drop the C-State folder into the Terminal window. You should now see the path of C-State folder pasted in the Terminal. Press Enter to navigate to that directory in Terminal
Linux users should know their stuff :)
Host a static server using one of the following ways:
Python 2.x: python –m SimpleHTTPServer
Python 3.x: python –m http.server
PHP: php –S localhost:8000
Node.js:
npm install -g http-server # install dependency
http-server -p 8000
Linux and MacOS installations come with Python preinstalled. You can check your Python version using the command python –V and use the appropriate command as listed above.
For a comprehensive list of ways to host static servers, you can check this page.
Once the server is hosted, you can access C-State by opening your browser and going to http://localhost:8000.
The following table lists the features of C-State as compared to other popular genome browsers.
| Feature | UCSC | IGV | WashU | JBrowse | CHROMATRA | CisGenome | EpiViz | visPIG |
|---|---|---|---|---|---|---|---|---|
| Simultaneous display of multiple user-specified loci | ||||||||
| Multiple feature tracks | ||||||||
| Concomitant display of gene expression data | ||||||||
| Comparison of user-generated and published datasets | ||||||||
| Versatile filters to allow selection of user defined patterns | ||||||||
| Interactive plotting | 5 | |||||||
| Modular and extensible | ||||||||
| OS agnostic | NA3 | |||||||
| Programming knowledge not required | ||||||||
| Prior dependencies not required | NA | 1 | NA | |||||
| Runs locally | 2 | 4 | ||||||
| Open source license |
- Feature available
- Feature unavaible
NA - Not applicable
Foot notes:
1 – Needs Java Run Time, which may not be preinstalled on some systems
2 – If using the JBrowse Desktop version
3 – Runs as a plugin for Galaxy Portal
4 – If Galaxy is installed as a local instance
5 – Plotting feature could not be tested on our data
Note: ChAsE is a recently published tool for interactive visualization of input regions. As opposed to C-State's gene-centric approach, ChAsE enables comparative analysis of epigenomic datasets via k-means clustering and heat maps.
Currently, C-State supports the following genome builds:
No. As C-State works with coordinate based information, any annotation or feature that can be described in BED format (e.g. exons, CpG islands, restriction sites, SNPs etc.) can be plotted as data tracks.
Any numerical value which indicates the relative expression level of a gene is accepted by C-State. For example, they could be raw read counts, RPKM, FPKM, CPM or TPM values from RNA-Seq experiments, or Normalized Log Ratios from microarray experiments. However, as expression values are compared across genes and cell types, we recommend using normalized values such as FPKM or CPM.
C-State was primarily designed for analysis and filtering of genes based on their chromatin signatures. Hence, it is fully functional even if no expression data is loaded. In fact, C-State can also handle data inputs where the expression of only some of the target genes is loaded.
C-State supports UCSC genome browser-type shading of features based on scores, which are in a range of 1 to 1000 (1000 being the best and 1 being the least score). Ensure that your feature scores are also in the same range. You can also disable the shading feature altogether by either removing your score information or by using a score of 0 for all features in your input files. In such cases, the score is displayed as 'NA' on mouse-hover.
C-State runs in all browsers which support HTML5 and ES2015 (the 2015 version of JavaScript). If you have updated your browser in the last one year, then your browser can run C-State. We have extensively developed and tested it in Google Chrome and hence recommend it, however we had no issues running it in latest versions of Mozilla Firefox and Microsoft Edge.
There are no arbitrary limitations imposed by C-State with respect to the number of genes, cell types or features to be plotted. However, the actual number that can be realistically plotted at one go is limited primarily by the total available RAM in the system. Also, plotting too many cell types or conditions on a single page (more than 8 in our opinion) can be aesthetically constraining.
This happens if some of the gene identifiers specified did not match any identifier in the genome (because of aliases etc.). C-State alerts the user in case some IDs could not be mapped. If none of the identifiers could be mapped, it mostly indicates that the chosen identifier type is wrong. C-State does not let the user proceed to next steps unless at least one gene identifier is successfully mapped.
This is a browser setting that can easily changed. For example, on Google Chrome, you can go to Settings -> Show Advanced Settings -> Downloads, and check the box that says "Ask where to save each file before downloading". On Mozilla Firefox, you can change this from Preferences -> General -> Downloads and select the radio button that says "Always ask me where to save files"
If C-State encounters an error in processing files, it flashes an alert. The timer spinning endlessly could indicate a system memory issue. In such a case close any other applications and reload C-State.
Last updated on 27th September, 2017.
Website developed by Saketh Saxena; Maintained and managed by Divya Tej Sowpati.
The intial prototype for C-State v0.1 was created by Saurabh Gaur.