Some frequently asked questions, and some hypothetical ones for fun
See the JBrowse_Configuration_Guide for more details
To setup jbrowse on your server, the basic idea is to download the zip file, drop the files in a folder on a web server, and run the setup.sh script.
You can do this for example by
Then you should be ready to go. Note that as long as setup.sh completes the “Installing Perl pre-requisites” phase of setup.sh you should be good to go (the other steps are not strictly needed)
Now open up your website e.g. http://localhost/jbrowse and you’ll already see that is is “setup” and you can see http://localhost/jbrowse/?data=sample_data/json/volvox has example data
Then you can simply run a small script to get your own genome loaded. For example, while you are in /var/www/html/jbrowse folder, then run
bin/prepare-refseqs.pl --fasta your_genome.fa
This will create a folder named “data”, and then visiting http://localhost/jbrowse/ will automatically load this “data” folder and youll see your genome.
Alternatively: you can try the JBrowse Docker container for a containerized deployment, or the JBrowse Desktop program to run without a web server (similar to IGV)
If you have JBrowse installed to your web folder and have run setup.sh, then you can download a FASTA file for your genome and run
bin/prepare-refseqs.pl --fasta yourfile.fasta
This will setup a “data” subfolder inside your jbrowse subdirectory with your genome prepared to view. You can investigate data/trackList.json and other files in this “data” folder to see what it generated.
Then just open up “http://yourserver.com/jbrowse” and it should be visible with the reference sequence as a track. By default, visiting “http://yourserver.com/jbrowse” will read files from that generated “data” folder. Other configurations are possible, but this is the easiest one.
The most common feature track to use is a GFF file
Using flatfile-to-json.pl is the easiest and most optimal way to load a GFF file for jbrowse
You can run
bin/flatfile-to-json.pl --gff myfile.gff --trackLabel trackLabel
Refresh your browser and your GFF file should be visible as a track. You may also experiment with adding –trackType CanvasFeatures, which is a newer track type that is more flexible to configure, or –type mRNA to only load mRNA features from the GFF, or other parameters that are listed in flatfile-to-json.pl –help.
When you set up a BAM file in jbrowse, the best way to do it is as follows
{
"storeClass" : "JBrowse/Store/SeqFeature/BAM",
"urlTemplate" : "mybam.bam",
"label" : "mybam",
"type" : "JBrowse/View/Track/Alignments2",
"key" : "mybam"
}
This basically just instructs JBrowse to fetch “mybam.bam” using the class that understands BAM data, “JBrowse/Store/SeqFeature/BAM”, and display it with the “Alignments2” tracktype. If this does not work please post to github or gmod-ajax@lists.sourceforge.net
Other notes
You can also add your config into a tracks.conf file instead of the trackList.json file, but you use a different format, for example
[tracks.mybam]
type=Alignments2
urlTemplate=mybam.bam
This is functionally equivalent to the trackList.json entry above
When you set up a BigWig file in jbrowse, the best way to do it is as follows
{
"storeClass" : "JBrowse/Store/SeqFeature/BigWig",
"urlTemplate" : "file.bw",
"label" : "mybw",
"type" : "JBrowse/View/Track/Wiggle/Density",
"key" : "mybw"
}
You could also add the –plot argument to add-bw-track.pl to use the XYPlot version of the bigwig track.
Alternatively, you can manually add that snippet into trackList.json, or add this alternative track conf form into your tracks.conf file
[tracks.mybw]
storeClass=JBrowse/Store/SeqFeature/BigWig
urlTemplate=file.bw
type=JBrowse/View/Track/Wiggle/Density
Notes:
This is where the training wheels come off
There is no add-vcf-track.pl, so you have to edit trackList.json yourself
First bgzip and tabix your vcf file
bgzip myfile.vcf
tabix -p vcf myfile.vcf.gz
If your VCF isn’t sorted for any reason, just use the GNU sort utility to sort it by chromosome and coordinate or get vcf-sort from vcftools
Now that your VCF is indexed, follow these steps
{
"label": "mytrack",
"urlTemplate": "myfile.vcf.gz",
"storeClass": "JBrowse/Store/SeqFeature/VCFTabix",
"type": "CanvasVariants"
}
That isn’t too bad right? All that add-bam-track.pl does is automate that for your similarly for BAMs, so now that you have edited the config file by hand you are ready to take on the world!
Note you could also add your track to tracks.conf
[tracks.mytrack]
urlTemplate=myfile.vcf.gz
storeClass=JBrowse/Store/SeqFeature/VCFTabix
type=CanvasVariants
Both of these configurations represent the same track.
A short trackList.json with an IndexedFasta reference sequence track would look like this
{
"tracks" : [
{
"label" : "refseqs",
"key": "Reference sequence",
"storeClass" : "JBrowse/Store/SeqFeature/IndexedFasta",
"urlTemplate" : "SOAPdenovo-genome.fa",
"useAsRefSeqStore" : true,
"type" : "Sequence"
}
],
"refSeqs" : "SOAPdenovo-genome.fa.fai"
}
You can see from this that a couple things are needed
With this setup, you do not need to have run prepare-refseqs.pl on a FASTA file. Instead you can simply use the “samtools faidx” program to index your fasta file in a data directory, and set trackList.json up in this format.
Most established HTTP servers such as Apache or nginx can work. You can just unpack JBrowse into your http web directory (/var/www/html, or htdocs, or similar) and follow the setup instructions http://jbrowse.org/code/JBrowse-1.12.1/docs/tutorial/.
Other configuration about your server will depend on your needs, but generally no special setup is needed for the http server, and there is no “server side” code used by JBrowse (cgi, php, etc).
Note: servers like “SimpleHTTPServer” from Python or “http-server” from NPM are generally not full featured enough to run all JBrowse features correctly (SimpleHTTPServer does not support Range queries, and http-server interprets tabix files incorrectly). RangeHTTPServer does pass the test suite for jbrowse though, so it should work for tests https://github.com/danvk/RangeHTTPServer (but not for compressed json files).
Also note: sometimes, people will download JBrowse and double click the html file and open up file:///c/myfolder/jbrowse/index.html in the webbrowser, however, running JBrowse like this is not recommended. This may appear to work for some limited cases but will fail on many others. You can use JBrowse Desktop if you want to use jbrowse locally without a webserver.
Since 1.12.4: The JBrowse “jb_run” script in 1.12.4 can be used to run JBrowse. It uses a custom express.js server tailored to serve all content headers directly.
Ignoring version numbers here, the difference is that one is a “release” version and one is a “development” version. If you are going to be modifying source code of JBrowse’s codebase, you probably want the development version.
The “release” version goes through a build step with minification so that loading the page takes less bandwidth. Different parts of the code go into different “layers”, but the main layer that includes most JBrowse source code and JBrowse libraries goes into src/dojo/dojo.js
The “development” version does not go through minification, and so it is amenable to editing the source code. You can also use a github clone for developing source code modifications. See the next section for running from a github clone #How_do_I_run_the_code_using_the_github_clone
Running code off github is fairly straightforward, and you can actually run it on your webserver without any build steps
git clone https://github.com/gmod/jbrowse
cd jbrowse
./setup.sh
Then move that directory into your web server root and you are ready to go.
Note: If you want to make your own custom build you can use
make -f build/Makefile release
Also also note: post 1.13.0 releases of JBrowse will automatically create a minified custom build when you run ./setup.sh if you have a -dev release or a github clone.
JBrowse uses both json and “.conf” files for configuration (the .conf is a custom format that is parsed on the client side), and both file types can contain the same types of information
Examples of JSON format
Examples of conf format
If you wonder why you would use the .conf format, some benefits include: easy to edit, easy to append to with text file tools, can define multi-line functions (JSON doesn’t allow newlines, so callbacks are specified on a single line).
All the config files are combined using a system of “includes” at runtime. The “order” behind loading all the config files is something like this:
Internally it uses 0-based coordinates
When you view the data (like in the view details popups, and in their position on the browser) it is 1-based.
The functions like feature.get(‘start’) would return 0-based.
There are a lot of differences!
Glyphs are a “unit” of drawing in a CanvasFeatures track. The glyph is just code that is responsible for drawing a feature on the screen.
Generate-names.pl will create a “search index” on, by default, the “names, IDs, and Alias” fields for tracks loaded with flatfile-to-json.pl or biodb-to-json.pl. It will not try to index ids from BAM files or bigwigs, but it does index names from VCF files too.
You can select specific tracks that you want to index with –tracks arguments to generate-names.pl. You can disable “autocomplete” by setting –completetionLimit 0 on generate-names.pl. You can “update” your search index by using –incremental
Also note: you can index additional fields of a GFF file with generate-names.pl by specifying the –nameAttributes flag to flatfile-to-json.pl. E.g. flatfile-to-json.pl –nameAttributes “my_custom_field,name,id”. The default value for this flag is “name,alias,id”. Note that the flag is passed to flatfile-to-json.pl and then these values are automatically indexed by generate-names.pl later.
The track “label” is more like the track ”identifier”, it should be unique! The key is actually more like the name that is displayed for the track. It might sound counter intuitive to have label and key this way. Key is not a required attribute, but label is. The label can be specified by –trackLabel on command line tools. The key can be specified by –key.
Some people don’t know this, but the box that shows your current location, e.g. “chr10:1..1000 (1.0 Kb)” is also a search box! You can search for things that generate-names.pl indexed here.
Also, the search index can be used to “link” to features, for example, if you construct a link such as http://localhost/jbrowse/?loc=GENE1234
Then the search index will resolve the location of that gene and jump to it automatically.
Try out jbrowse_elasticsearch, it is still experimental but it allows this. Or, implement your own JBrowse REST names API. The default generate-names.pl is not built for searching full text descriptions.
By default, the scripts will output to a subdirectory called “data” in the jbrowse folder
You can control that output with most scripts using the –out parameter. This enables you to have “multiple data directories”.
Once the data directories are ready then use the URL bar to select which data directory to use with ?data=my_data_dir e.g.
http://mysite.org/jbrowse/?data=data1 http://mysite.org/jbrowse/?data=data2
Note that with Apollo, you can output the data directories to some given directory and add the directories via the user interface.
You can also setup the “dataset selector” see #What is the dataset selector and JBrowse_Configuration_Guide#Dataset_Selector
The dataset selector is a dropdown that can list all the genomes that are in your jbrowse instance
To configure the dataset selector, set a dataset_id inside your trackList.json or tracks.conf on your data directory, and then in jbrowse.conf, add a list of all your datasets with the dataset_ids that you listed in the genome’s data directory.
See http://gmod.org/wiki/JBrowse_Configuration_Guide#Dataset_Selector
If you don’t like the names in the “Name” or “ID” column of your GFF, and you instead want to use some other field as the name to be displayed, then you can add this to your trackList.json
"style": {"label": "my_custom_field"}
Note: you can also index “my_custom_field” with generate-names.pl too by supplying my_custom_field to the –nameAttributes argument from flatfile-to-json.pl. After loading it from flatfile-to-json.pl in that manner, it will be indexed by generate-names.pl.
Also note: you can make the description a custom field too in a similar way
"style": {"description": "my_custom_description_field"}
In CanvasFeatures, this is done with the style->color parameter. The style->color parameter can be a function, so for example, if you have a track like this in trackList.json
{
"label": "test",
"type": "CanvasFeatures",
"storeClass": "JBrowse/Store/SeqFeature/NCList",
"style" : {
"className" : "feature"
}
}
Then you can add a color like this
{
"label": "test",
"type": "CanvasFeatures",
"storeClass": "JBrowse/Store/SeqFeature/NCList",
"style": {
"color": "function(feature) { return 'red'; }",
"className": "feature"
}
}
This will make your features red. You can also hardcode a color instead of a callback
"style": {
"color": "red"
}
It can be dynamic too though
"style": {
"color": "function(feature) { return feature.get('score')>50 ?'blue':'red'; }"
}
The color can be a name or rgb(…) or hsl(…). rgba works too
Note: if you get a very complex function, consider putting it in a separate functions.conf file and include it, see config guide “Including external functions in trackList.json”
Note: with HTMLFeatures, it is similar, but you have to use hooks->modify instead of style->color.
Also note: you cannot create the functions via the UI in “Edit config”. You have to specify the callback functions via the config files themselves.
You can add custom info to the “About this track” dialog boxes by specifying metadata for the track in trackList.json, e.g.
"metadata": {"legend": "Red means this, green means that"}
This will create a box called “legend” inside the “About this track” box.
All the “metadata” fields are added to the “About this track” dialog box, and it can also support HTML, e.g
"metadata": {"randominfo": "<p style='color:green'>Green text</p>", "otherrandominfo": "<p style='color:red'>Red text</p>"}
You can also use the fmtDetailMeta and fmtFieldMeta callback functions to change pre-existing values http://gmod.org/wiki/JBrowse_Configuration_Guide#Additional_customizations_to_the_pop-up_boxes
There are many ways to do so.
Those are just a couple examples
More straightforwardly, basically any field that exists for the feature (e.g. all the things in column 9 for a GFF) will be added to the default “View details” boxes, so it you add more details to the GFF3 column 9, then your popups will have more information.
Typically this is done using a plugin. You can make your plugin add new menu items. See the RegexSequenceSearch plugin for an example https://github.com/GMOD/jbrowse/blob/master/plugins/RegexSequenceSearch/js/main.js#L23-L30
You can also create a new “menu” i.e. (default says file, view, help) and you can extend it to use (file, view, tools, help) or similar. The GMOD/Apollo codebase does this https://github.com/GMOD/Apollo/blob/master/client/apollo/js/main.js#L368-L379
In general, you would want to make a new “track type”, you can override _trackMenuOptions on your custom track type. See https://github.com/cmdcolin/gccontent/blob/master/js/View/Track/GCContentXY.js for example
You can edit the menuTemplate parameters on the trackList.json.
Note: when you add 1 method to menuTemplate, it overwrites 1 of the default elements in the default right-click menu. This is sort of a bug. If you want to “add” your method to the list, then create a “blank” menuTemplate items for the View details, etc. See JBrowse_Configuration_Guide#Customizing_Right-click_Context_Menus for more details.
Each feature is based on what is called a feature Model. The JBrowse/Model/SimpleFeature is an example of this, and it allows you to call feature.get(‘variable_of_interest’) on many different variables
Common types of requests
Note that if you have multiple values for a value in column 9 (specified by comma separated values), then feature.get(‘your_value’) will return an array (except for for Note which turns into Note, Note2, Note3, etc)
For reference JBrowse/Model/SimpleFeature is a widely used feature class. The BAM features for example use a different model because they must be smaller and quicker to operate at speed (they are called Lazy features, so not all their info is evaluated unless it is needed on-demand). BAM features also have things like feature.get(‘seq’), to get the sequence of the alignments, but normal features like genes do not have feature.get(‘seq’)
There are several config variables which you can define in any of your config files (trackList.json for example can have config items like this, just put it outside of the track section of tracklist.json, or under a [general] section of tracks.conf) as a comma separated list of track labels
The easiest way to “embed jbrowse” on another page would be to use an iframe to link to the jbrowse instance of choice e.g. http://gmod.org/wiki/JBrowse_Configuration_Guide#Embedded_mode
Note that it is also possible in theory (and sometimes in practice) to embed JBrowse without an iframe. For example, you can take the index.html of JBrowse and customize it for your purposes. You can for example change the GenomeBrowser div to not have 100% width and height, and instead make it have width 1000px and height 800px for example and put it in a specific part of your webpage.
The problem with embedding without iframe is typically that you can easily run into issues with “CSS collisions” where the CSS from your specific page collides with jbrowse elements, of CSS from jbrowse elements collides with your outside page.
Yes. It is sort of a “hidden setting”, but you can actually change colors on the Sequence/Alignments2/SNPCoverage track using CSS
See css/sequence.css:
/* colors for bases must be specified as hex or rgb/hsl strings, no named colors such as 'red' */
.base_n {
background-color: #C6C6C6;
}
.base_a {
background-color: #00BF00;
}
.base_c {
background-color: #4747ff;
}
.base_t {
background-color: #f00;
}
.base_g {
background-color: #d5bb04;
}
.base_reference {
background-color: #a33;
}
.base_deletion {
background-color: #999;
}
If you change those, it will be reflected in your Sequence/Alignments2/SNPCoverage track.
The .base_reference is the background grey of the SNPCoverage plot, and if you actually add to the list and make .base_skip you can make the intron grey change color SNPCoverage.
If you use sequenceserver, you create a custom JBrowse link by creating a links.rb file and running
sequenceserver -D database_dir -r links.rb
The links.rb is then a file that gets included by sequenceserver to generate new links to external resources automatically. In our case, we parse all the HSPs (high scoring pairs) from the hits from BLAST and string them together into a single feature (e.g. by taking the min and max of all hsps)
Note: The links.rb passes us the accession and hsps variables implicitely. The notation for the query start/end and subject start/end are shortened as qend, send, sstart, and qstart
We use the addFeatures URL parameter of JBrowse to make the features appear. For reference, see http://gmod.org/wiki/JBrowse_Configuration_Guide#Controlling_JBrowse_with_the_URL_Query_String
require 'json'
module SequenceServer
module Links
def jbrowse
qstart = hsps.map(&:qstart).min
sstart = hsps.map(&:sstart).min
qend = hsps.map(&:qend).max
send = hsps.map(&:send).max
first_hit_start = hsps.map(&:sstart).at(0)
first_hit_end = hsps.map(&:send).at(0)
my_features = ERB::Util.url_encode(JSON.generate([{
:seq_id => accession,
:start => sstart,
:end => send,
:type => "match",
:subfeatures => hsps.map {
|hsp| {
:start => hsp.send < hsp.sstart ? hsp.send : hsp.sstart,
:end => hsp.send < hsp.sstart ? hsp.sstart : hsp.send,
:type => "match_part"
}
}
}]))
my_track = ERB::Util.url_encode(JSON.generate([
{
:label => "BLAST",
:key => "BLAST hits",
:type => "JBrowse/View/Track/CanvasFeatures",
:store => "url",
:glyph => "JBrowse/View/FeatureGlyph/Segments"
}
]))
url = "http://yourwebsite.com/jbrowse/" \
"?loc=#{accession}:#{first_hit_start-500}..#{first_hit_start+500}" \
"&addFeatures=#{my_features}" \
"&addTracks=#{my_track}" \
"&tracks=BLAST" \
"&highlight=#{accession}:#{first_hit_start}..#{first_hit_end}"
{
:order => 2,
:title => 'JBrowse',
:url => url,
:icon => 'fa-external-link'
}
end
end
end
Important note: if you have multiple organisms, then you should encode your FASTA files to contain the organism name, e.g.
>Human-chr1
ACCGAATCAGCTATCGA...
This is because BLAST does not tell you which database your hits come from.
Then you must modify your links.rb use this information by parsing the “accession” variable.
require 'json'
module SequenceServer
module Links
def jbrowse
qstart = hsps.map(&:qstart).min
sstart = hsps.map(&:sstart).min
qend = hsps.map(&:qend).max
send = hsps.map(&:send).max
first_hit_start = hsps.map(&:sstart).at(0)
first_hit_end = hsps.map(&:send).at(0)
organism = accession.partition('-').first
sequence_id = accession.partition('-').last
my_features = ERB::Util.url_encode(JSON.generate([{
:seq_id => sequence_id,
:start => sstart,
:end => send,
:type => "match",
:subfeatures => hsps.map {
|hsp| {
:start => hsp.send < hsp.sstart ? hsp.send : hsp.sstart,
:end => hsp.send < hsp.sstart ? hsp.sstart : hsp.send,
:type => "match_part"
}
}
}]))
my_track = ERB::Util.url_encode(JSON.generate([
{
:label => "BLAST",
:key => "BLAST hits",
:type => "JBrowse/View/Track/CanvasFeatures",
:store => "url",
:glyph => "JBrowse/View/FeatureGlyph/Segments"
}
]))
url = "http://yourwebsite.com/jbrowse/" \
"?data=#{organism}" \
"&loc=#{sequence_id}:#{first_hit_start-500}..#{first_hit_start+500}" \
"&addFeatures=#{my_features}" \
"&addTracks=#{my_track}" \
"&tracks=BLAST" \
"&highlight=#{accession}:#{first_hit_start}..#{first_hit_end}"
{
:order => 2,
:title => 'JBrowse',
:url => url,
:icon => 'fa-external-link'
}
end
end
end
Another important note: you should BLAST the actual genome sequence if you are linking to genomic coordinates of JBrowse. If you BLAST a CDS database, you may need to load CDS sequences into JBrowse, or translate the CDS coordinates into genomic coordinates
Also also note: if you are running sequenceserver through apache e.g. with Passenger Phusion, then you can modify the config.ru to have a “require ‘./links.rb’” line and keep the links.rb in that same directory as config.ru e.g.
require 'sequenceserver'
require './links.rb'
SequenceServer.init(:config_file => ".sequenceserver.conf")
run SequenceServer
Yes! Try the jbrowse desktop versions, built with electron!
The Windows and OSX versions are easy to use, and all you need is to open your fasta file (ideally: indexed fasta).
You can also open BAM tracks, BigWig, VCF.gz, GFF3, BED, BigBed, and more!
Yes! The packages are not hosted on cpan, but you can install them from github using cpanm
cpanm git://github.com/GMOD/jbrowse.git
This will install jbrowse scripts such as prepare-refseqs.pl and flatfile-to-json.pl to, commonly, a folder like ~/perl5/bin if you are using local::lib, or simply a system folder like /usr/local/bin if using sudo.
There are a couple scripts that don’t work with this such as maker2jbrowse, but it is otherwise fine to install the perl scripts this way.
Note: you might also use –notest option to avoid testing all dependencies
className refers to a CSS class for your features.
If you are using CanvasFeatures, this is an unused artifact.
If you are using HTMLFeatures, then you can add custom CSS to make your feature have a custom class. Note that the “subfeatureClasses” is a related variable: it is a CSS class for subfeatures.
By default, it would just use the “exon” class for exons or whatnot, but subfeatureClasses allows you to create a map e.g.
"subfeatureClasses": {"exon": "myCustomExonCSSClass"}
The most reliable way to do this is to use gff3sort from https://github.com/billzt/gff3sort
See http://biorxiv.org/content/early/2017/06/04/145938 for a description of their algorithm
Note that you can try and use GNU sort (sort -k1,1 -k4,4n) or genometools (gt gff3 -sortlines) but these both have problems where it will place child features behind the parent features in the GFF
In JBrowse 1.14, the problem of child features being behind their parents was fixed so the full GFF3Sort algorithm from @billzt is now not necessary and a simple GNU sort does work.
JBrowse 1.12+ allow opening FASTA files directly in the browser or via JBrowse Desktop. Indexed FASTA is however much more efficient as it does not require being read into memory.
To create an Indexed FASTA, install samtools and run
samtools faidx yourfile.fa
This will create a file called yourfile.fa.fai. When you want to open up your own sequence file in JBrowse, you can then use the “Open sequence file” option, and drag and drop both the .fa and the .fai in the file area. JBrowse will understand that these are to be used together, and will open it.
Note: you can also open unindexed FASTA, but it requires parsing the whole FASTA up front, so this is slow and memory intensive with the current setup. Indexed FASTA is quite efficient though.
If the BAM and VCF files you have are large, the BAM index or TABIX index files can become large as well. Since the indexes must be fully downloaded before any of the data can be displayed, you can break your files up by chromosomes, and use {refseq} in a urlTemplate to break it up into manageable chunks.
E.g.
"urlTemplate": "myfile_{refseq}.bam"
That would search for myfile_chr1.bam and myfile_chr1.bam.bai when you open that track while browsing chr1
Yes, you can configure one. See this section http://gmod.org/wiki/JBrowse_Configuration_Guide#Configure_a_Loading_Page
You could also listen for other milestone events to trigger other loading events (see DebugEvents plugin for details)
Try using –completionLimit 0 with the command. It will disable autocompletion but still allow you to search exact matches.
Note that you can use generate-names with –completionLimit 20 on some tracks and then generate-names with both –incremental and –completionLimit 0 flags on very information dense tracks.
The hierarchical data selector can support multiple levels of drop down categories. To use this, set the “category” variable on your track, and use a “/” to represent a subcategory. Use multiple / for multiple subcategories.
Example
{
"category": "ParentCatgory / DiseaseBAM",
"label": "myTrack",
"storeClass": "JBrowse/Store/SeqFeature/BAM",
"type": "Alignments2";
},
{
"category": "ParentCatgory / NonDiseaseBAM",
"label": "myTrack2",
"storeClass": "JBrowse/Store/SeqFeature/BAM",
"type": "Alignments2";
}
In tracks.conf form
[tracks.myTrack]
category=ParentCategory / DiseaseBAM
type=Alignments2
storeClass=JBrowse/Store/SeqFeature/BAM
You can set the following config
collapseCategories=ParentCategory1/ChildCategory,ParentCategory2/ChildCategory
etc. to your jbrowse.conf. This is a comma separated list (don’t include spaces around the slashes though). Remember, don’t quote the values in the jbrowse.conf file :)
You can use the –type argument for flatfile-to-json.pl
E.g.
flatfile-to-json.pl --type mRNA —gff mygff.gff
This will only load mRNAs from the GFF. Additionally, if you want to filter on the source column of the GFF, you can augment the –type argument with an extra formatted parameter for source –type mRNA:augustus
The –type argument can also be a commas separated list of filters like this.
prepare-refseqs.pl accepts a –sizes parameter, which takes a “chrom.sizes” file which is just a tab separated file with two columns, refseq names and their lengths
This let’s you view the genome and the features on the genome without loading the sequence data.
After version 1.13 which introduced webpack to JBrowse, then you must re-run “setup.sh” after adding or removing a plugin.
Also, you must use the JBrowse-1.x.x-dev.zip instead of just JBrowse-1.x.x.zip
To install a JBrowse plugin, generally the easiest thing to do is to put the code in the plugins directory (e.g. clone the repo to plugins/GCContent or plugins/SashimiPlot), and then just say this in your jbrowse.conf (or tracks.conf
[GENERAL]
plugins += GCContent
plugins += SashimiPlot
Equivalently, in jbrowse_conf.json (or trackList.json), that would mean just having
"plugins": ["GCContent", "SashimiPlot"]
Essentially, you are just giving the configuration an array of folder names to look for in the plugins directory
Alternatively, you can also add the “name” and “location” of the plugin too.
"plugins": [{
"name": "GCContent",
"location": "plugins/gccontent"
}]
Here “plugins.GCContent” means that the “name” of your plugin is “GCContent”. This is slightly important, because the “name” is used for the purposes of the “namespace” that the plugin has i.e., when you specify a track with the “type”: “GCContent/View/Track/GCContentXY” in trackList.json, then that means the “name” of the plugin should be GCContent, not lowercase gccontent.
Note that the above config would be equivalent to this in the jbrowse.conf format
[ plugins.GCContent ]
location = plugins/gccontent
Again, you don’t need to use this format if the name of the directory is as expected e.g. plugins+=GCContent looks for a folder named GCContent
Note: “location” is a relative URL to the jbrowse root directory. Normally when a plugin is successfully installed, you will get a console.log message from the plugin saying that it has started up (that depends on the plugin but most jbrowse plugins just do that by convention)
Another note: you can add plugins declarations to trackList.json or tracks.conf instead of jbrowse_conf.json or jbrowse.conf too. In tracks.conf just put it at the top of the file or underneath a line that says [GENERAL]
Final note: it is best to only include plugins in one file, e.g. put all the includes in jbrowse.conf or all configs in tracks.conf, not mix them in different config files.
If your web service doesn’t exactly match the requirements for the JBrowse REST API tracks, then you can create your own “store class” as a plugin. This basically just requires one thing:
bin/new-plugin.pl MyPlugin
Then, simply make a dojo class (using “dojo declare”) in your plugin that implements a “getFeatures” function. The getFeatures function receives a query object with query.start, query.end, query.ref e.g. chr1 along with 3 callbacks: featureCallback, finishCallback, and errorCallback. If there is an error, than call the error callback obviously. Otherwise, for each feature that you want to display, call featureCallback with that (use JBrowse/Model/SimpleFeature to represent the feature). When you are out of features for the query region, call finishCallback.
Check out the genes store class from the myvariantviewer plugin for an example of a simple custom adaptor https://github.com/cmdcolin/myvariantviewer/blob/master/js/Store/SeqFeature/Genes.js
A JBrowse plugin can do a wide variety of things. Some common use-cases would be
Other things that are extensible include
One cool thing is that when you create a plugin, you can simply reference it textually in your config file e.g. trackList.json, and then JBrowse will load the “class” from your plugin and initialize it
Make sure to review this link describing dojo declare, the way dojo declares new object types http://dojotoolkit.org/reference-guide/1.10/dojo/_base/declare.html
Also review http://dojotoolkit.org/documentation/tutorials/1.9/modules/
Understanding “*dojo declare*” and *asynchronous module definition (AMD)* will help you understand the “preamble” on the top of every file that jbrowse uses, and which you can use in your plugin
In version 1.13 and forward, the plugin system also relies on re-building JBrowse using webpack. This actually allows you to use ES6 javascript and node js modules in your browser code. Therefore, you might want to review webpack, babel, and node js module systems
Let’s walk through a simple plugin with the goal of adding something to the track menu (e.g. where the “About this track” and “Save track data” options are)
To do this, we can use object- oriented principles to “inherit” from some existing track type like CanvasFeatures and then extend its functionality by overriding the functions in a new track type
We can inherit a new track type by using the “define” function to include the dependencies needed in a file, and they are listed in an array at the top of your file.
First initialize a new plugin using
bin/new-plugin.pl MyPlugin
Then edit a new file, say plugins/MyPlugin/js/MyTrack.js
define( ["dojo/_base/declare", "JBrowse/View/Track/CanvasFeatures"],
function(declare,CanvasFeatures) {
return declare(CanvasFeatures, {
_trackMenuOptions: function() {
var opts=this.inherited(arguments); //call the parent classes function
opts.push( // add an extra menu item to the array returned from parent class function
{
label: "Custom item",
type: 'dijit/CheckedMenuItem',
onClick: function(event) {
console.log('Clicked');
},
iconClass: "dijitIconPackage"
}
);
return opts;
}
});
}
);
Code listing 1. an example custom track type, plugin/MyPlugin/js/MyTrack.js, that adds an extra track menu item
After this, we will have the plugin directory structure like this
jbrowse/plugins/MyPlugin
jbrowse/plugins/MyPlugin/js
jbrowse/plugins/MyPlugin/js/main.js
jbrowse/plugins/MyPlugin/js/MyTrack.js
The bin/new-plugin.pl helps create the skeleton main.js
Then we can use our new plugin to a config file like jbrowse_conf.json as “plugins”: [“MyPlugin”]
Then edit the trackList.json for an existing track and change `“type”: “CanvasFeatures”` to `“type”: “MyPlugin/MyTrack”`.
That will tell jbrowse to load the MyTrack class from your plugin instead of the normal CanvasFeatures class.
Inspect your setup.log
If for example it says
version.c:30:16: fatal error: db.h: No such file or directory
Then run
sudo apt-get install libdb-dev
Then re-run setup.sh
If you see errors for XML::Parser
sudo apt-get install libexpat1-dev
Then re-run setup.sh
Also make sure you use “./setup.sh” or “bash setup.sh”, do not use “sh setup.sh”
This error is often due to some system issues about compiling libraries like libpng, and for all intents and purposes can be ignored, as it is only used in wig-to-json.pl and this is superceded by directly reading BigWig files (no conversion step needed)
You can follow these instructions for how to setup a BigWig file if needed http://gmod.org/wiki/JBrowse_FAQ#How_do_I_set_up_a_BigWig_file.3F
This message normally means that jbrowse is setup but a genome hasn’t been loaded or located correctly
You can continue by running
bin/prepare-refseqs.pl --fasta yourfile.fa
Then reload the page and your genome should be available.
Note: If the red box on the “Congratulations page” shows a different message than just 404 on seq/refSeqs.json, then report the error to github or the mailing list with as much detail about your setup as possible.
This can normally be fixed by deleting ~/.cpanm
It may be due to conflict between jbrowse’s own cpanm and your system cpanm, but it should not be too problematic.
Generally deleting ~/.cpanm is harmless, it is a “build” directory (generally ~/perl5 is the local::lib directory, and in jbrowse’s case, it actually uses an alternate local::lib directory named extlib inside the jbrowse directory to ensure ease-of-install)
From what we have seen, the “Integer overflow error” sometimes appears on BigWig tracks when your webserver is not configured correctly. It seems to be due to errors with a “reverse proxy” or something not forwarding the data properly.
Therefore, it is most likely not due to corrupted bigwig files or jbrowse bugs, but more probably, due to your server’s configuration.
This error basically says the “search function” from generate-names.pl isn’t working. You can try a couple things to fix the error
Note if there are continued troubles, you can try an alternative search engine, such as jbrowse_elasticsearch (an experimental plugin) https://github.com/cmdcolin/jbrowse_elasticsearch/
If you get an error similar to this:
_Argument "-" isn't numeric in addition (+) at /Library/WebServer/Documents/scbrowse/JBrowse-1.12.0/bin/../src/perl5/Bio/JBrowse/FeatureStream/GFF3_LowLevel.pm line 32, <$f> line 44611._
Make sure your GFF3 is tab delimited
Increase maxFeatureScreenDensity to a higher value. This value is by default 0.5 but if you allow a higher “density” of features, set it to 6 for example and the message should disappear.
Several things can happen to cause this (generally on VCF of BAM tracks)
Some things to check:
defaultTracks=mytrack1,mytrack2 not
defaultTracks="mytrack1,mytrack2"Also note: specifically with regards to the “defaultTracks” parameter, defaultTracks is overridden by the users cookies and the &tracks= parameter in the URL, so to test whether defaultTracks works, clear you cookies and visit without &tracks in the url. Use alwaysOnTracks or forceTracks if you want to have it turn on despite cookies/URL.
If you get the error such as this
Too many open files opening bucket log /path/to/your/data/names/00f/6.json.log at perl5/lib/perl5/Bio/JBrowse/HashStore.pm line 197, <$fh> line 85.
Then try increasing number of files available
ulimit -n 1000
The default can sometimes be as low as 256 (view with ulimit -a)
This is normally due to a module called mime_magic being enabled on your Apache server. Two options for fixing this are 1. disable mime_magic or
Your server is misconfigured for VCF.GZ files, and this can be due to it thinking that it should set “Content-Encoding: gzip” on the your .vcf.gz files. Your webserver should actually NOT put “Content-Encoding: gzip” on your VCF.GZ files. If you think you have this problem, can try using “curl -I” to view the headers that your server is putting on your VCF.GZ file.
Note: this issue can be confusing to research about, especially the “Content-Encoding: gzip” issue, because most information on the web typically says that it should put “Content-Encoding: gzip” on gzip data, and this header implies that the client should decompress the content itself, however JBrowse does not want this to happen because VCF.gz files are a special type of gzip, specifically, bgzip, so it is manually decompressed by JBrowse javascript code.
If you have a very dense track with many features, JBrowse might decide to hide the labels to save space, but you can force them to display again by adding this to your trackList.json
"style":{"labelScale": 0.01}
This says that the label will be displayed when the zoom level is greater than 0.01 regardless of how many features are there. The value 0.01 is measured in pixels per base pair, at max zoom level, there are 25 pixels per base pair for example, and when zoomed out farther, each base takes up less space, hence 0.01 will display the feature names always if you are reasonably zoomed in.
You can also change maxHeight to a larger value to make the track taller and see more features.
This normally means some javascript code for handling the track has crashed. Check your javascript console for clues on how to fix it. Add a github issue if it represents a real bug!
Note: you should use the “-dev” packages for debugging, i.e. JBrowse-1.11.6-dev.zip as opposed to for example JBrowse-1.11.6.zip, because the -dev package contains “un-minified” source code and more readable javascript console messages
If your GFF does not follow this structure
gene->mRNA->exon+CDS
Then you need to add extra configuration
Specifically, if it is “transcript” instead of “mRNA” (which is common for Ensembl GFF for example), then you must set
"transcriptType": "transcript"
Also, if you only have “exon” and no “CDS”, then you need to set the subParts config (the default settings assumes that both exons and CDS exist, so if there are only exons, like in a cufflinks output file, then you need this)
"subParts": "exon"
If your GFF does not include UTR, but the UTR can be “implied” from the difference between the exon and CDS boundaries, then you can use this on your track type to enable them
"impliedUTRs": true
If your GFF file has features with this structure
match -> match_part
This only has two levels, you might consider just setting the “Segments” glyph
"glyph": "JBrowse/View/FeatureGlyph/Segments"
The segments glyph accepts all subfeatures, so match and match_part structure is fine.
Note: the tips above only apply for CanvasFeatures tracks
HTMLFeatures generally load data at the “transcript” level. This means that they should be loaded with something similar to –type mRNA when using flatfile-to-json.pl in order to see the transcript subfeatures
flatfile-to-json.pl --type mRNA --gff your_genes.gff --trackLabel MyTrack
This means that it loads the features where mRNA would be in column 3 of your GFF. If it was an Ensembl GFF, you might use instead –type transcript
Note that this also loses the information about the “parent” gene feature however, so it might be worth loading an additional track at the gene level like
flatfile-to-json.pl --type gene --gff your_genes.gff
This track will not display the transcript and exon subfeatures, but instead just show a box where the gene is, so this is commonly called a “gene spans” track
Note: If you would like a track that displays with the transcript subfeatures, you can use the CanvasFeatures type track (i.e. load with flatfile-to-json.pl –type gene –trackType CanvasFeatures …)
Your GFF should use proper ID and Parent relations. Your subfeatures do not need to themselves have IDs if they have no further subfeatures, but they must have a Parent pointing to the Parent’s ID
Note that it should be spelled Parent, not PARENT
If you get the error “Building and installing legacy bam-to-json.pl support (superseded by direct BAM tracks) … failed. See setup.log file for error messages. If you really need bam-to-json.pl (most users don’t), try reading the Bio-SamTools troubleshooting guide at https://metacpan.org/source/LDS/Bio-SamTools-1.33/README for help getting Bio::DB::Sam installed.”
Then note:
This can happen if the chromosome names from your track don’t match the names from your reference genome.
Try and make sure the chromosome names from your evidence tracks match the chromosome names from the reference genome fasta.
Examples of error messages
Check that
You can use code such as this
<iframe id="jbrowse_iframe" src="/jbrowse/" scrolling="no" style="width:100%;height:800px;"></iframe>
<script>
//https://stackoverflow.com/questions/2090551/parse-query-string-in-javascript
function getQueryVariable(variable) {
var query = window.location.search.substring(1);
var vars = query.split('&');
for (var i = 0; i < vars.length; i++) {
var pair = vars[i].split('=');
if (decodeURIComponent(pair[0]) == variable) {
return decodeURIComponent(pair[1]);
}
}
}
// subscribe to jbrowse movements inside of the iframe and update parent page url
var datadir = getQueryVariable('data');
var iframe = document.getElementById('jbrowse_iframe');
iframe.addEventListener('load', function() {
var JBrowse = iframe.contentWindow.JBrowse;
JBrowse.subscribe( '/jbrowse/v1/n/navigate', function(obj) {
var shareURL = JBrowse.makeCurrentViewURL();
var parser = new URL(shareURL);
window.history.replaceState( {}, "", parser.search );
});
});
// pass the parameters from the parent page into the iframe
iframe.src = iframe.src + window.location.search;
</script>
With this setup, you can pass URL parameters from the URL of the parent page e.g. http://localhost/parent_page/?data=mydata&loc=chr1:1..10000 and it will forward those URL params to the jbrowse instance (located at http://localhost/jbrowse) and the URL will be autoupdated when you change locations
Yes! See http://gmod.org/wiki/JBrowse_Configuration_Guide#Rendering_high_resolution_screenshots_using_PhantomJS for an example
Since flatfile-to-json.pl does not accept GTF, you can convert your GTF to GFF3. Tools like gffread or gtf2gff3.pl are available
The gffread tool is packaged with cufflinks so simply install cufflinks, then you can run
gffread -E merged.gtf -o- > merged.gff3
Yes! Upload jbrowse to a gh-pages branch on a github repo, and also put a .nojekyll file in the root directory.
This bypasses the normal jekyll parser of github and allows jbrowse to load https://github.com/blog/572-bypassing-jekyll-on-github-pages
In general, using normal commands like flatfile-to-json, prepare-refseqs, etc work fine though. See setup.sh for how the volvox sample data combines using biodb-to-json and other techniques.
Yes you can!
The styles on “CanvasFeatures” include normal, compact, and collapse
By default, compact divides the height of glyphs by 4, so if you make the height of your features smaller with style->height then when you set compact it will be ultra compact.
Yes! Try setting style.featureScale to a very small number like 0.0000000001 (but greater than 0)
Yes, try out the SashimiPlot plugin! https://github.com/cmdcolin/sashimiplot
It dynamically calculates the splicing coverage of a track or uses junctions.bed files for junctions
Yes, the GCContent plugin will calculate the GCContent from your sequence data automatically. See https://github.com/cmdcolin/gccontent
It works fairly well on mid-size genomes. If you have very large megabase scale assemblies, then you might consider pre-calculating the GCContent.
Yes, the GWASViewer plugin does this. https://github.com/cmdcolin/gwasviewer/
Yep! The regular alignments track types (e.g. JBrowse/View/Track/Alignments2) supports RNA-seq and will show spliced alignments.
Also, there are two special options for RNA-seq that can help decipher the reads.
Long reads from platforms like nanopore and pacbio pose some challenges but will work if it is in BAM format. The JBrowse 1.12.3 release also includes an optimization, cacheMismatches, to enhance speed on long read tracks. This must be enabled manually in the config at the moment.
You can make a custom plugin to do this. The “multibigwig” plugin is an example of this https://github.com/cmdcolin/multibigwig
Also note: with the third option, you can make it so that your BAM track has a bigwig when zoomed out, but then shows the reads when zoomed in. Any CanvasFeatures track can use a bigwig for summary histograms. The Alignments2 volvox-sorted.bam track is an example of this. See http://gmod.org/wiki/JBrowse_Configuration_Guide#Configuring_Summary_Histograms
Yes. You can set the config variable view.maxPxPerBp to a higher value. To increase, try setting this in jbrowse.conf
view.maxPxPerBp=50
Note sometimes the ”translations” will appear wrong at high zoom levels, so don’t depend on this for the protein translations
By default, the max zoom level is 25, so setting it to 50 makes you able to zoom in twice as much.
The pos_color and neg_color config variables for BigWig tracks accept callback functions. The phytozome browser has good examples of this with the VISTA plot tool