Don't get me wrong, I'm strongly convinced that Galaxy is a powerful tool to manipulate data interactively.
But as a trained bioinformatician who have a deep knowledge of the linux command line, do you consider that it is only a tool for the "biologist/end-user or do you actually use galaxy for your every day data-manipulations (why?).
We're about to install galaxy here, and some colleagues believe that it will be used from 'A' to 'Z' to make our NGS analysis. Why not ? But why should I spend some time in front of my web browser when I can write a shell-script (and save it into git).
OH: Using Galaxy with big data requires local installation, which requires competence, but with such competence you don't need to use Galaxy
1) We are hoping to improve the documentation of our workflows, so that it is clear who ran what script, what version of what script, and what the output was. This (we hope) should improve reproducibility and we are anticipating it will help to address reviewer comments that bemoan the "data was analyzed using local scripts" sentence many of us may have written into our methods. This is probably the main reason a bioinformatics group might begin using Galaxy. Of course, there are other solutions to this -- but Galaxy is just one of those solutions.
2) Allows non-computational investigators in the group to begin learning NGS data analysis without first taking 4-6 months to learn Linux at the command line. But not all groups will have this need -- we happen to have this need. So, it will probably be group dependent.
And a third reason, I have to add, the Galaxy development team are an excellent group of professionals to work with.
Do you remember "Works on my machine" certification program ? Galaxy provides an environment for reproducibility of a workflow. If I take your shell script, I need to make sure I have the same libraries, environmental variables, paths, etc. But I can take your Galaxy workflow, assuming you have a reasonably standard Galaxy instance, and expect it to run it on my data without thinking about compatibility.
What about an attempt to reproduce certain workflow from 10 years old paper? Even if you have the source code, it's almost 100% chance it is not going to work. Galaxy is a substitute of a virtual machine.
I think the main reason for a power user like yourself to use it would be that you could actually contribute to Galaxy itself. In that way your power tools could come available to all these non power users and you could help to clean up that overfilled tools shed by removing things you see in there that are not as good as other things. (according to ISMB talks there literally are thousand of tools in that shed).
Command-line savvy power users are not its target audience, and I don't see any reason why you and I should be using it. Just like most GUIs, it fills an important niche for non-computational types, but is limiting to a power user.
In spite of my CLI knowledge, I use and appreciate galaxy (the JGI implementation) mainly for the workflows.
As a more general thought, I really do think that being a bash and command-line expert is a real advantage as a bioinformaticist, however when something good happens in the GUI world I won't throw it away. For instance, who does still use pine or lynx? ;-)
For me the main reason to #usegalaxy is in terms of training students and dealing with collaborators. First, galaxy allows you to teach bioinformatics separately from computing (e.g. UNIX, programming). Second, Galaxy allows you to reduce the time consuming back-and-forth of communicating methods/results between you and students/collaborators. I estimate that somewhere upwards of 60% of a project can be simply communicating methods and results - sharing histories reduces this drastically. Third, by providing constraints on what a user can do, it prevents a naive user from tearing down a machine/cluster, so you spend less time f-ing around on system administration with a newbie, which puts many of them off from being self-sufficient. Fourth, it establishes a best practice environment for learning bioinformatics - eg. Galaxy allows you to teach concepts of workflows easily & code organization/testing/documentation when developing local tools. Finally, by having a new users start with Galaxy and learn its limits for themselves, they quickly become motivated to take their training wheels off and learn how ride for themselves. In sum, Galaxy is worth adopting for you to be able to get non-trained biologists to be able to help themselves, which gives you more time to focus on the real science, rather than being a bioinfortechnician. As I say to many people: "Use galaxy, it's out of this world"
Other points on this thread are key but I'll add my experience as a core facility manager. Apart from the obvious advantage of allowing inexperienced computer users a GUI environment, an insta-gui for new command-line scripts and programs (that automatically plays nice with other programs) enables rapid prototyping. Thus the user can play with options that allows them to import results easily into other programs and workflows which may even be set up by the core beforehand.
Also the library feature is useful, allowing shared data to selected groups.
For me Galaxy is mainly used to do some manual jobs like intersect my regions of interest with genome tracks from UCSC. In bioinformatics you really need to control your data by manually looking into it time-to-time, thats why GUI tools are useful. I do this during downstream functional analysis, and I believe it is the easiest way in most cases.
However, I do not use it to pipe data from raw reads to filtering/mapping and so on (despite there is a nice workflow system implemented in Galaxy), as I feel running shell scripts is more stable.
I do get benefits by using Galaxy by doing some training with out any bioinformatics experience and computer basic skills at all, especially when I could not get access to the high powerful computer but my personal computer. At least I got a general whole picture of pipeline in my head from beginning like reads quality control until SNP calling. '
Now, since I get access to powerful computer and know some operations of Linux and sequence softwares, pre-trainning experience by Galaxy really give me some ideas at least that I know how to do for the next steps using some protocols provided by many guys from sequence communities and make me clear that what is the wrong or right of the out put files like...
In my experience with some instances of Galaxy, often while using it you will find interesting tools that you want to adopt for your command line workflows. In those cases, then you can go to the toolshed and clone the repository with mercurial. The tool in question might be a script that has been wrapped for usage in Galaxy, in which case it the repository will also have an XML wrapper file. Hopefully, the script is well documented so that you can use it locally.