Skip to content

Victorian-Bioinformatics-Consortium/degust

Repository files navigation

This repository is deprecated

You should now use https://github.com/drpowell/degust

Degust

  • Visualise RNA-seq differential expression data.
  • Perform your own DGE analysis, or use the inbuilt server to analyse from your own "counts" file.

Access a public web service running Degust.

View a short video of the interface in use.

Read a summary on the Degust home page.

Example Screenshot

Degust screenshot

FAQ

How can a gene have zero counts for all samples but has a non-zero fold-change?

This can happen when using the backend of Degust. Degust uses voom to perform the expression analysis. Voom adds a small constant (0.5), to each count, normalizes for library size, then takes the log. This means when you have a count of 0 across all samples, but different library sizes, it is possible to compute a non-zero fold-change.

We recommend setting Min read count on the configuration page to a small value, say 10.

What is the Min read count setting?

This is the minimum number of reads required in at least one sample to keep the gene in the analysis. That is, a given gene is omitted if the number of reads across all samples is below this setting.

Chrome dies with an "Aw, Snap" error when trying to download the table

This appears to be a problem with Chrome when there are many (thousands) of genes in the table. We suggest using Firefox if this happens.

I don't see an option to display an MDS plot

The MDS plot is only available when you have included "count" columns

How is the MDS plot calculated

  • First genes that don't pass the "FDR cut-off" or "abs log FC" filters are ignored. Using these filters is "cheating" when doing an MDS plot to look at replicate clustering
  • The remaining genes have the counts for each replicate log-transformed.
  • The genes are then ranked by decreasing variance. That is, the most variable genes are "at the top"
  • Then the top "Skip genes" are ignored.
  • And the next "Num genes" are selected.
  • These selected genes are used to compute an MDS (or PCA) plot

Installation

If you do not want to use the public Degust installation, you may install your own.

You first need to grab a copy of Degust.

    git clone git@github.com:Victorian-Bioinformatics-Consortium/degust

Degust can be installed in two ways:

  1. Perform your own DGE analysis, and use only the web frontend from Degust
  2. Install the frontend and back-end software to perform analysis and visualise the results.

Frontend installation only

To use the frontend visualisation, you will need to have done your own DGE analysis with a tool like edgeR or voom. You will need CSV file contain a line per gene, and the following columns:

  • ID - containing a unique identifier for each gene (required)
  • Adjusted p-value - The adjusted p-value (FDR or similar) for that gene (required)
  • Log intensity for each condition - Used to compute the log fold-change (required)
  • Average intensity across the conditions - Used for the MA-plot (required)
  • Gene info - Arbitrary information columns to display in the gene list table (optional)
  • Read counts - Read counts for each replicate, only used for display purposes (optional)

The simplest approach is to download degust.py then run it with your csv file as a parameter. This will create a single HTML page that you view or share. Run ``degust.py --help` to find the parameters to specify the column names for your CSV.

Full installation

Run tests locally

There are some javascript tests which can be run locally (or with travis). Build the js:

./build-tests.sh

Then you can either run the tests in your browser (navigate to http://localhost:8000/)

(cd tests/js/ ; python -mSimpleHTTPServer)

Or, if you have phantomjs installed you can run the tests from the command line: ./test-js.sh

Contributing

Feel free to contribute with pull requests, bug reports or enhancement suggestions.

Development

To build

For building from sources, you will need nodejs and the following modules.

npm install -g browserify
npm install -g clean-css
npm install hbsfy@1.3
npm install handlebars-runtime
npm install uglify-js
npm install coffeeify

# Builds files into build/ for deployment
./build.sh prod

For development

This will watch the js & coffeescript files and rebuild CoffeeScript as needed.

npm install -g watchify

./build.sh dev
./build-watchify.sh &
(cd build ; ../server.py)

To build the degust.py script for embedding csv into an html file for local

./build-embed.sh local

It is also useful to access the pages with "debug=1" (eg. http://vicbioinformatics.com/degust/compare.html?code=example&debug=1) which enables extra debug logging to the console.

To build with the analysis back-end

The above production build only includes the front-end. To also build the back-end you can use the following. (The haskell library requirements are not well documented yet.)

./build.sh prod-server

Requirements:

  • Python
  • CoffeeScript
  • R and the following libraries
    • limma
    • edgeR
  • GHC 6.12 or later, and the following libraries:
    • pureMD5 >= 2.1
    • json >= 0.7
    • regex-pcre >= 0.94
    • hamlet >= 1.1
    • shakespeare-text >= 1.1
    • strict-io >= 0.2
    • lens >= 3.9

The resulting build/ directory can then be installed as a CGI site.

These specific steps are known to work on an ubuntu 14.04 install.

Troubleshooting

  • The directories "tmp/", "cached/" and "user-files/" under the CGI directory must be writable by the web-server user
  • Any runtime errors relating to R will be logged in the directory "tmp/" under the CGI directory
  • If your R libraries are not installed in the default location, you may need to edit r-json.hs and modify the setting for R_LIBS_SITE

License

Degust is released under the GPL v3 (or later) license, see COPYING.txt