nodalpoint.org - BioRuby - Comments

prove your point

Neil — Tue, 14 Jun 2005 07:54:53 -0400

We are a tough audience here at Nodal. On the one hand, geeky enthusiasm is what we live for. On the other - if you storm in full of big promises, you'd sure better live up to them. Because we have a zero tolerance policy for bullshit.

Do not make the fatal mistake of assuming that bioinformatics is merely a branch of computer science. Because when I read claims like "analysis of almost every type of file", then I see a bunch of clustering algorithms which construct trees with nodes, I wonder how that applies to everyday bioinformatic tasks like annotating genomes or batch-processing PDB files.

Richness of data is the hallmark of biological information. I've no doubt that this software is interesting and important, but before telling me that it can do everything, take some time to understand biological problems.

complearn

cilibrar — Tue, 14 Jun 2005 05:42:38 -0400

Sure, there are many reasons details in my research papers on my homepage: http://www.cwi.nl/~cilibrar/

In summary the main reasons are
a) it's easier, usually
b) it's more robust, usually

I would like to fix the Mac OSX compile/link problems. Can anybody let me have an account for a day or two on their Mac to fix this problem? Please email me at cilibrar@gmail.com if you are up for helping.

Best regards,

Rudi

Exciting new site ?

Greg — Sun, 12 Jun 2005 23:36:36 -0400

Nodalpoint will be celebrating it's 5th anniversary in a few months. I think "exciting" rather than "exciting new" site might be a better way to put it.

So can you offer us any reason why a compression based approach is better than [insert favorite machine learning technique applied to biology here] ?

Compiling on OS X

alf — Sat, 04 Jun 2005 11:22:24 -0400

I had no luck compiling this on OS X. Things like malloc.h and -export-dynamic throw up easily fixable errors, but I wasn't able to find a way round an undefined symbol error later on.

Tools

Anonymous — Wed, 11 Aug 2004 13:54:13 -0400

You are both right at some extent. But I used to use nice and easy packages that did pretty much what I neede years ago. I had a Mac and Macvector. We did comparisons, restrictions analyses and we produced figures for publications. Now we have all these genomes, huge databases and new tools (blast did not exist at the time). But most of the Open Source tools and packages lack even simple and user friendly sequence editors! Not to mention graphical based tools to make analysis of complex and huge datasets (Ensembl being an outanding exeption) . The idea with the package was to take away the "mystique" with all obscure command line commands (in EMBOSS who knows what octanol or fuzztran means?? only by the name?). To bring these new tools to "normal" biologists. Now that people is installing the system we have a bigger user community and we discuss the pros and cons with the systems/programs. This is something that many programmers should love: input from the users and not only from other computer geeks. Otherwise the two communities evolves in different directions. People produces Excellent programs that not many biologists use and this has a price: Both the Staden and EMBOSS core developers main institutes have no funds to continue. With a bigger user community this should have being impossible.
And YES a bigger efford in Documentations is also a MUST. We are preparing several documents based on the questions and comments from the increasing number of people that install our package.
Regards

PS. EMBOSS.GUI is also a pretty piece of work and we have installed that GUI in many computers worldwide. We are thinking in incorporating it in the package.

Working is good

Neil — Tue, 10 Aug 2004 02:42:00 -0400

Mmm, have I touched a nerve :-)

I agree that ideally, things should "just work", so they can be put to use. I also agree documentation is important. But I think we have to be realistic about the nature of most open-source software. It is written largely on a voluntary, unpaid basis by people in their spare time out of the goodness of their hearts. They expect some effort from the user to get it working and rightly so.

Take as an example the excellent EMBOSS GUI, which is my new favourite EMBOSS interface. It does pretty much "just work", provided you have set up a working web server and CGI bin (i.e. you have root access). A "casual user" might well complain about this things, but these things are not for a casual user. They are for a sysadmin, who runs a central server available to anyone at the institution who wants it.

I am in no way elitist - I just don't think that a lot of these pre-packaged "bioinformatics for dummies" ideas are very helpful in tackling the big issue - which is bringing the concept of the "information age" in biology to the traditional experimental biologist.

Tools -- and their use

chris — Mon, 09 Aug 2004 22:25:38 -0400

The idea with bioinformatics tools is to use them to solve biological problems not to use months of work to get the packages working.

Agreed. I'd rather not dick around for months getting stuff to work. However, I know what I'm doing -- some of the time -- so I don't always need the learning curve. I'm not advocating either ivory towers or mystic priesthoods for computational biology, mind you. But complex tools require complex knowledge. Expecting stuff to "just work" ignores both the intricacy and experimental nature -- to say nothing of dubious quality -- of much of the software available.
There seems to be a disconcerting trend amongst many "bench" scientists that a result from a computer analysis is (i) correct and (ii) the product of a few mouse clicks. Being both a bench and computational person, I really don't understand this.

[...] many developers are incredible bad in producing documentation.

Perhaps more effort should be spent on documentation than packaging then?

Devil

Anonymous — Mon, 09 Aug 2004 18:16:48 -0400

Stupidity has nothing to do with this. The problem is that many developers are incredible bad in producing documentation.
1-Have you tried to install wwwblast following the pour and contradictory documentation that follows with it?
2-Or have you tried to find on the EMBOSS documentation where and how to configure the databases needed to run the programs? Look for e.g. at BioKnoppix: EMBOSS is there but nothing works! This MacOSX package installs EMBOSS and all databases needed.
3- The idea with bioinformatics tools is to use them to solve biological problems not to use months of work to get the packages working.
4- The package is not perfect and a new release with many enhancements and pre-installed GUIs will be available.
5- The web-pages will also be changed ;-)

Devil's advocate

Neil — Wed, 07 Jul 2004 03:43:36 -0400

I welcome initiatives like this - but let me play devil's advocate for a while :-)

In my experience, people who are too stupid to install a package are too stupid to use it. Does making the install easier help such people?
Do we need a lot of individuals doing easy installs on a lot of workstations? Or one individual taking care of a number of workstations and providing help for other users?
Perhaps I've become a command line nazi, but there's a place for GUIs and I'm not sure it's EMBOSS. Most EMBOSS packages are as easy as typing "program_name sequence_file". It actually takes far longer typing into the individual fields of a GUI and the GUI is only any use if it intelligently processes the program output. Which embossRunner may do, for all I know.

Still, it's good to encourage people to make more use of their computers and ebioinformatics.org seems to do that. Really nasty site design though ;-)