Training materials

This page contains instructions and data files for interactive training exercises.

You can also download a PDF slideshow designed to accompany these tasks, written for the Helsinki EMBRACE/ENFIN workshop.

Task 1 — running jobs and analysing the data

For this task, you will need a spreadsheet application such as Excel. Others such as Numbers or OpenOffice will work just as well, but the instructions below are specific to Excel 2007, so you may have to figure out what the equivalent commands are called in other applications.

You will also need the following two data files: query.txt and ref.txt. They contain 10 UniProt accessions each. Save both files to your desktop.

Go to the FuncNet web interface and start a new session using these two files as your query and reference proteins.

When the session is complete, you will see a page of results. Click the ‘download results in CSV format’ link. Open the CSV file directly in Excel if you are given this option, or else, save it to your desktop and open it manually.

Select all the cells containing p-values — i.e. all the predictions from individual predictors, from column D onwards. Click ‘Conditional Formatting’ -> ‘New Rule’ and create a rule to highlight all p-values less than or equal to 0.05 (e.g. give them a red background). The exact way to do this depends which version of Excel you have. This allows you to highlight statistically significant predictions, just like the web interface does, but lets you choose the p-value cutoff yourself.

Once you’ve done this, use ‘Conditional Formatting’ -> ‘Data bars’ on the FuncNet Score (column C) to apply a nice visual indication of the relative overall score for each pair. (This is not possible in older versions of Excel.)

Task 2 — using the Java client

For this task, you will need a reasonably recent (5+) version of the Java runtime environment. A full JDK with compiler is not required, as you will be using a pre-compiled binary distribution. You will also need the two protein list files query.txt and ref.txt from the previous task.

First, download the JAR file containing the software, and save it to your desktop, alongside the two text files. This is quite a large archive (13MB) because it includes Sun’s web services framework (Metro) and various other dependencies. The actual FuncNet-specific code in there is less than 50KB.

Open up a command-line window (i.e. cmd.exe in Windows, Terminal in OS X, or xterm/gnome-terminal/konsole in Linux) and change to the directory where you saved the JAR file.

Type the following to check that Java is working okay:

java -cp example-clients-dist-1.0.3.jar eu.funcnet.clients.jaxb.FrontEndClient

You should see a usage message. If you get an error at this stage (or at any stage!) please ask for help.

Assuming everything worked okay, type the following to submit the same query to FuncNet as you did in the previous task:

java -cp example-clients-dist-1.0.3.jar eu.funcnet.clients.jaxb.FrontEndClient query.txt ref.txt YOUR_EMAIL_ADDRESS false

… substituting YOUR_EMAIL_ADDRESS with your own address. This is just for session tracking and ownership purposes.

You will get some status messages indicating the progress of your job. After about five minutes you should see some scores. How does the score output from the command-line client differ from the web interface?

The example-clients package also comes with clients that let you send jobs to the individual FuncNet predictors directly. We will try this with GECO which uses a database of correlated gene expression patterns, and is usually very fast. To send the same query directly to GECO, type:

java -cp example-clients-dist-1.0.3.jar eu.funcnet.clients.jaxb.GecoClient query.txt ref.txt

You don’t need to supply your email address here as the predictors have no concept of a ‘session’ with an owner and a changing status. Instead, all processing is done during a single HTTP request.