Replay Analyses. The graphical client allows you to reload the parameters from any previous analysis you've run, and to replay it.
User namespaces.. User namespaces allow you to view public modules and your own, and protect your private modules from view by others
Search engine. Search engine allows you to search all documents on your GenePattern server, including task descriptions, documentation, scripts, job parameters, and output files. The search page on the website describes the syntax for multiple terms, exclusion, etc.
Other new features:
GenePattern pipeline environment home page is better organized.
Updated tutorial and documentation.
CPU and Operating System task parameters are now enforced.
<R> allows simple execution of tasks written as R scripts.
You can run a task directly from the web client, without creating a one-step pipeline.
You now identify yourself to the web, Java, and R clients in order to gain access to your private tasks. Private tasks are now hidden from other users, except on the GenePattern home page. The username you choose when identifying your tasks should not have any punctuation other than "@" and ".".
Installation and system configuration issues:
If your browser is configured to block popups, you won't be able to email pipeline results, stop running pipelines, or see various pipeline error messages displayed. In Netscape Navigator 7.1, this is controlled via Edit->Preferences->Privacy & Security->Popup Windows.
If a single computer has multiple instances of GenePattern running, they must not be allowed to share a single HSQL instance. Doing so would cause some jobs to be run multiple times, and job results would be randomly retrievable from one server or the other. HSQL will quietly die if started a second time on the same port, but the clients that connect to a single instance will all be "happy", though not for long.
If you are using a version of R which you cannot modify (because it is a publicly-shared version and you don't have appropriate privilege), you can have it load the GenePattern library by setting the environment variable R_LIBS=<GenePattern install directory>/R/library in your autoexec.bat, .cshrc, .bashrc or other shell startup file. R will then load from its usual location, but will also search for and find the GenePattern library from your installation.
Pre-installed tasks have their documentation in Adobe Acrobat PDF file format. You will need to have the free Acrobat Reader installed on your computer in order to view this documentation. It can be downloaded from http://www.adobe.com/reader.
Under Linux, if you do not have write permissions into the R installation directories your install will complete but will note some errors during install. The server will be unable to execute pipelines unless you reinstall with the appropriate privileges or follow the directions (above) for using a version of R which you cannot modify.
Under Mac OS X, you must have write permission for the /Applications folder in order to install R.
Under Mac OS X, both Netscape Navigator and Mozilla have problems accessing "http://localhost:8080" if the computer is not connected to the Internet. If you find yourself in this situation, change your server name from "localhost" to "127.0.0.1".
If you install the server on a non-default Tomcat port (not 8080), the GenePattern Client installer will not automatically select this port for you to connect to. You must change it manually during installation.
To install under Mandrake Linux 9.2, you must install the binutils, libstdc++, gcc, and libxml packages. Use rpmdrake to search for and install them easily. You should also install the xpdf application if you don't already have a PDF viewer.
If you encounter OutOfMemory errors in the server or client when attempting to load or operate on large datasets, you will need to enlarge the memory space for the GenePattern server or client. To do this, edit the file in the GenePattern directory called StartGenePatternServer.lax (server) or in the GenePatternClient directory called GenePattern_Graphical_Environment.lax. In either of these files look for the entries noted below and increase these values by doubling up to the max memory size of the machine you are using.
"Prompt when run" for pipeline parameter selections to be made at pipeline execution time, does not work for pipelines run in the browser environment, only for those run from the Java or R clients.
If you edit the pipeline R code and later regenerate the pipeline using the pipeline designer, your code changes will be overwritten without warning.
If you edit the R pipeline code and have an error in your code, the error message reported will be "Error: syntax error, execution halted." No mention is made of the line on which the error occurs.
If you rename a task, pipelines that refer to the old task name are not updated to reflect the new name. Similarly, if you rename or delete parameters from a task, pipelines that refer to that task are not updated and will fail.
Pipeline input files with spaces in their names may give "file not found" errors. If this happens, use DOS' "dir /x" command to get the 8.3 version of the directory and filename and use that instead of the long filename. If you are using a Unix-based platform, you may need to quote the filename parameters on the command line definition.
The pipeline designer functions best when browser page caching is disabled. In Netscape Navigator 7.0 and 7.1, go to Edit->Preferences->Advanced->Cache, set the cache size to zero and clear the cache.
Zipping a pipeline for sharing does not recursively zip the associated tasks used by the pipeline. Zip them separately if you need to.
Your GenePattern username is used as the sender email address for pipeline results. Some SMTP mail servers require that this be a valid, local email address. Using simply a username without "@domain.name" may not work. If it fails you will see a pop-up browser window allowing you to edit the sender address when a pipeline completes and notification was requested.
Graphical client issues:
The File system view does not show drive roots (ie. other drive letters) under Windows.
You may see performance decrease if you load project directories that contain a very large number of files in them.
The lower left pane does not show file size (aka "available") for prediction results files, and possibly for others as well.
When errors or other messages appear, their windows may be hidden underneath the main client window. If you click in the client window and get no response, you may need to switch (Windows Alt-Tab, Mac Cmd-Tab) to that window and close it.
Files deleted from the Project Dir node aren't deleted from Data Objects view.
If you include spaces in the name of an input file, the job will produce an error.
Output files that contain a semicolon in their name can't be viewed.
Viewing .html files on platforms other than Windows and Mac OS X is hardwired to "netscape". As a workaround you can create a soft link from your browser to netscape (e.g. ln -s /usr/bin/mozilla /usr/bin/netscape)
Task creation, execution, and sharing issues:
GenePattern does not distinguish between two users with same-named tasks. If one user creates a task with the same name as an existing one owned by a different user, it will overwrite it.
Avoid providing input files with spaces in the filenames, other than "Program Files".
On servers running Microsoft Windows, the following input filenames will not work, regardless of the extension they are given: con, prn, aux, clock$, nul, com, com1, com2, com3, com4, lpt1, lpt2, lpt3.
If a task is renamed to match the name of some other existing task, you will not be warned before overwriting the existing task.
If you created any pipelines under GenePattern 1.0, you will get an error when you attempt to run them in GenePattern 1.1. Open and save them in the pipeline designer to rewrite the pipeline and it will work again.
PCAViewer: Install Sun's Java 3D for a 3-dimensional viewer. The viewer will work without it, but will not provide 3-D visualization.
SVM: Confidence values are not computed and are set to the value 1.
PreprocessDataset (and other perl modules) will produce incorrect output if fed files with line endings that are different from the host operating system's line endings. For example, submitting a file with windows line endings to PreprocessDataset on a Linux GenePattern server will produce incorrect results.
HierarchicalClustering will produce either an atr file if you clustered by samples (columns) or a gtr file if you clustered by genes (rows).
PreprocessDataset (and other perl modules) will produce incorrect output if fed files with Mac line endings. Use the new ConvertLineEndings task to convert them prior to processing.
Scripting environment issues:
The sample R code shown at the bottom of taskWrapperGenerator.jsp, with "askUser()" prompts for each parameter, will only work correctly under Windows. Other platforms do not have support in R for prompting the user with this modal dialog. Filling in the parameters when the method is invoked will work correctly on all platforms.