Sequence Tube Maps

A JavaScript module for the visualization of genomic sequence graphs. It automatically generates a "tube map"-like visualization of sequence graphs which have been created with vg.

Link to working demo: https://vgteam.github.io/sequenceTubeMap/

Biological Background

Recent scientific advances have lead to a huge increase in the amount of available genomic sequence information. In the past this sequence information consisted of a single reference sequence, which can be relatively easily visualized in a linear way. Today we often know multiple variants of a particular DNA sequence. These could be sequences from different individuals of the same species, but also homologous (= having shared ancestry) sequences from different species. The differences between the individual sequences are called polymorphisms and can range in size from variations of a single base pair to variations involving long stretches of DNA. These polymorphisms are a key focus point for all kinds of sequence analysis, since analyzing the differences between sequences and correlating them to possible differences in phenotype allows to make conclusions about the function of the analyzed sequence.

Graph data structures allow the encoding of multiple related sequences in a single data structure. The intention is to simplify the comparison of multiple sequences by making it easy to find the sequences' similarities and differences. There are a number of approaches (and file formats) for formally encoding variants of genomic sequences and their relationships in the form of graphs. Unfortunately it is often difficult to visualize these graphs in a way which conveys the complex information yet is easy to understand.

Functionality

The purpose of this module is to generate visual representations of genomic sequence graphs. The visualization aims to display the information about all sequence variants in an intuitive way and as elegantly as possible.

Genomic sequence graphs consist of nodes and paths:

• A Node represents a specific sequence of bases. The length of this sequence determines the node's width in the graphical display.
• A Path connects multiple nodes. Each path represents one of the sequences underlying the graph data structure and its walk along multiple nodes.

This simple example shows two paths along three nodes:

Since both paths connect the same nodes, their sequences are identical (and the three nodes could actually be merged into a single one). If the two sequences would differ somewhere in the middle, this would result in the following image:

The way genomic sequences change in living organisms can lead to subsequences being inverted. For these cases, instead of creating two different nodes, a single node is traversed in two different directions:

The sequenceTubeMap module uses these elements as building blocks and automatically lays out and draws visualizations of graphs which are a lot bigger and more complicated.

There already exist various JavaScript tools for the visualization of graphs (see D3.js force field graphs or the hierarchy-based approach by Graphviz). These tools are great at displaying typical graphs as they are usually defined. But these regular graphs consist of nodes and edges instead of paths and have significant differences compared to genomic sequence graphs. Regular graphs have edges connecting two nodes each and not continuous paths connecting multiple nodes sequentially, nor do their nodes have a forward or backward orientation. We therefore need a specialized solution for displaying genomic sequence graphs (nevertheless the sequenceTubeMap module uses D3.js for the actual drawing of svg graphics after calculating the coordinates of the various components).

Usage

Online Version: Explore Without Installing Anything

The easiest way to have a look at some graph visualizations is to check out the online demo at https://vgteam.github.io/sequenceTubeMap/. There you can play with visualizations from a few different data sets as well as look at some examples showcasing different structural features of variation graphs. You can even provide your own vg-generated data as an input (limited to small file sizes only).

Run Sequence Tube Maps On Your Own

If you are using vg and want visualize the graphs it generates, the online version is limited to small file sizes. For visualizing bigger data sets you can run Sequence Tube Maps on your own. You can either run Tube Maps completely on your local (Linux) machine or use your local browser to access Tube Maps running on any other (Linux) machine you have access to.

(Previously we provided a docker image at https://hub.docker.com/r/wolfib/sequencetubemap/, which contained the build of this repo as well as a vg executable for data preprocessing and extraction. We now recommend a different installation approach.)

Prerequisites

• The NodeJS version specified in the .nvmrc file, which as of this writing is 18.7.0. Other several other NodeJS versions will work, or at least mostly work, but only this version is tested. THis version of NodeJS can be installed on most systems with nvm.
• NPM or yarn. NPM comes included in most NodeJS installations. Ubuntu packages it as a separate npm package.
• vg (vg can be tricky to compile. If you run into problems, there are docker images for vg at https://github.com/vgteam/vg_docker.)

The directory containing the vg executable needs to be added to your environment path:

PATH=/<your_path_to_folder_with_vg>:$PATH

Installation

• Clone the repo:

  git clone https://github.com/vgteam/sequenceTubeMap.git

• Switch to the sequenceTubeMap folder
• Install npm dependencies:

  yarn install

  or

  npm install

• Build the frontend:

  yarn build

  or

  npm run build

Execution

• Start the node server:

  yarn serve

  or

  npm run serve

• If the node server is running on your local machine, open a browser tab and go to localhost:3001.
• If the node server is running on a different machine, open a local browser tab and go to the server's URL on port 3001 http://<your server's IP or URL>:3001/. If you cannot access the server's port 3001 from the browser, instead of configuring firewall rules etc., it's probably easiest to set up an SSH tunnel.

ssh -N -L 3001:localhost:3001 <your username>@<your server>

Setting Up a Visualization

The application comes with pre-set demos that you can use to learn the tool's visual language and basic features.

To set up a custom visualization of particular files, you will need to configure a set of "tracks" describing the files you want to visualize, using the "Configure Tracks" dialog in "custom (mounted files)" mode. For information on how to do this, click on the "?" help button, or read the help documentation online.

Adding Your Own Data

• The vg files you want to visualize need to contain haplotype/path info. Generating visualizations for the graph itself only is not supported. In addition to the haplotype graph, you can optionally visualize aligned reads from a gam file.
• Your data needs to be indexed by vg. To generate an index of your vg file, go to the sequenceTubeMap/scripts/ directory and run

  ./prepare_vg.sh <vg_file>

  <vg_file> is the file name of your vg file including path information. If there are .vcf.gz and .vcf.gz.tbi files next to your .vg, they will be used to generate a GBWT index of haplotypes from the VCF. In this case, the .vg file must contain alt paths, from the -a option of vg construct.

• To generate an index of your gam file (optional, you can view vg only too):

  ./prepare_gam.sh <gam_file>

  <gam_file> is the file name of your gam file including path information.

• The output files will be generated in the same folder as the original files. To tell Sequence Tube Maps this location, edit sequenceTubeMpas/src/config.json and modify the entry for dataPath:

  "dataPath": "<path to my data folder>/",

  If you want to use a relative path, this path should be relative to the sequenceTubeMaps/ folder.

• restart the server and choose custom (mounted files) from the data dropdown in the UI to be able to pick from the files in your data folder.

Preparing subgraphs in advance

The sequenceTubeMap will fetch the necessary data when a region is queried. That can sometimes up to 10-20 seconds. If you already know of regions/subgraphs that you will be looking at, you can pre-fetch the data in advance. This will save some time during the interactive visualization, especially if there are a lot of regions to visualize.

The net result needs to be one or more chunk directories on disk, referenced from a BED file.

To generate each chunk, you can use the prepare_chunks.sh script. You ought to run it from the directory containing your input files and where your output chunks will be stored (i.e. the dataPath in sequenceTubeMaps/src/config.json), which defaults to the exampleData directory in the repo.

For example:

cd exampleData/
../scripts/prepare_chunks.sh -x mygraph.xg -h mygraph.gbwt -r chr1:1-100 -d 'Region A' -o chunk-chr1-1-100 -g mygam1.gam -g mygam2.gam >> mychunks.bed
../scripts/prepare_chunks.sh -x mygraph.xg -h mygraph.gbwt -r chr1:101-200 -d 'Region B' -o chunk-chr1-100-200 -g mygam1.gam -g mygam2.gam >> mychunks.bed

The BED file linking to the chunks has two additional nonstandard columns:

• a description of the region (column 4)
• the path to the output directory of the chunk, chunk-chr1-1-100 in the example above, (column 5).

chr1    1   100 Region A    chunk-chr1-1-100
chr1    101 200 Region B    chunk-chr2-101-200

Note each column is seperated by tabs

This BED file needs to be in the dataPath directory, or it can be hosted on the web along with its chunk directories and accessed via URL.

If you want certain nodes of the graph to be colored, place the node names to be colored in a nodeColors.tsv file, with a node name on each line, within output directory of the chunk. When rendered, these specified nodes will be colored differently than other nodes.

You can use prepare_chunks.sh script to generate this additional nodeColors.tsv by adding an additional option. Here is an example:

cd exampleData/
../scripts/prepare_chunks.sh -x mygraph.xg -h mygraph.gbwt -r chr1:1-100 -d 'Region A' -o chunk-chr1-1-100 -g mygam1.gam -g mygam2.gam -n "1 2 3" >> mychunks.bed

Adding this additional n flag will allow a string space delimited input of node names which will be outputted to nodeColors.tsv.

1
2
3

Pre-made subgraphs

You may want to look at a graph that has already been extracted from a larger graph. To support this, there is a prepare_local_chunk.sh script, which takes a subgraph rather than a full graph. It supports most of the options that prepare_chunks.sh does, with the notable exception of haplotype files. It assumes that the graph represents some region along some reference path that is present in the graph, and expects that region to be provided with the -r option. It assumes that path names in the subgraph don't use subregion suffixes (bracket-enclosed numbers). The path name used in the region should exactly match the name of one of the paths in the graph.

prepare_local_chunk.sh also accepts .gaf files, which will automatically be converted into a gam file using vg convert.

For example, you can run it like:

cd exampleData/
../scripts/prepare_local_chunk.sh -x subgraph.gbz -r chr5:1023911-1025911 -g subgraph_reads.gam -g other_sample_reads.gam -g another_sample_reads.gaf -o subgraph1 >> subgraphs.bed

Your graph can be a .vg, .xg, .gfa, or any other graph format understood by vg, but it must be in the same node ID space as your reads, and the script does not check this for you! In particular, indexing a GFA graph and mapping to it with vg giraffe can result in the original GFA nodes being cut into manageable pieces and assigned new numbers in the graph that the reads actually are aligned to, meaning the original GFA won't work here. You can check your reads against your graph with vg validate subgraph.gfa --gam subgraph_reads.gam. If your read alignments look completely absurd and jump all over the place, this is likely the problem.

If the original subgraph file does not remain in place under the configured dataPath and accessible by the tube map, errors may occur complaining that it couldn't be accessed when the tube map attempts to list ist contained paths.

The net result will be that you can select the BED file, select the region it specifies, and view a precomputed view of the subgraph, with coordinates computed assuming it covers the region provided to prepare_local_chunk.sh.

A note on naming node IDs when using .gfa files: VG keeps node IDs the same when all node names are strictly positive integers. However, node IDs are renamed upon encountering string-named nodes. Renaming begins at the first encounter of a string-named node, using the highest integer encountered so far (+1), or 1 if the first node is string-named in the GFA file. Future nodes are renamed in a +1 manner regardless of their datatype.

Here's an example of a rename:

Original -> Renamed
3 -> 3
1 -> 1
five -> 4
7 -> 5
four -> 6

Development Mode

The build/serve pipeline can only produce minified code, which can be difficult to debug. In development, you should instead use:

  yarn start

or

  npm run start

This will use React's development mode server to serve the frontend, and run the backend in a separate process, behind React's proxy. Local ports 3000 (or set a different SERVER_PORT in .env) and 3001 must both be free.

Running in this mode allows the application to produce human-readable stack traces when something goes wrong in the browser.

Running Tests

For interactive development, you can use:

  yarn test

or

  npm run test

This will start the tests in a watching mode, where files that are changed will prompt apparently-dependent tests to rerun. Note that this only looks for changes versus the currently checked-out Git commit; if you have committed your changes, you cannot test them this way. On Mac, it also requires that the watchman package be installed, because it needs to watch the jillions of files in node_modules for changes.

If you want to run all the tests, you can run:

  yarn test -- --watchAll=false

or

  npm run test -- --watchAll=false

You can also set the environment variable CI=true, or look sufficiently like a kind of CI environment known to reach-scripts.

If you want to run just a single test, based on its describe or it name argument, you can do something like:

  npm run test -- --watchAll=false -t "can retrieve the list of mounted xg files"

Running Prettier Formatter

In order to format all .js and .css files you can run:

npm run format

Currently, this repo currently uses Prettier's default options, including double quotes and 2 space tab width for JS.

License

Copyright (c) 2018 Wolfgang Beyer, licensed under the MIT License.